100% found this document useful (42 votes)

429 views

EBook Etextbook PDF For Applied Regression Analysis and Other Multivariable Methods 5Th Edition PDF Docx Kindle Full Chapter

ebook

Uploaded by

jeff.tinker162

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

100% found this document useful (42 votes)

429 views

EBook Etextbook PDF For Applied Regression Analysis and Other Multivariable Methods 5Th Edition PDF Docx Kindle Full Chapter

ebook

Uploaded by

jeff.tinker162

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

You are on page 1/ 62

(eTextbook PDF) for Applied

Regression Analysis and Other

Multivariable Methods 5th Edition
Visit to download the full and correct content document:
https://ebookmass.com/product/etextbook-pdf-for-applied-regression-analysis-and-oth
er-multivariable-methods-5th-edition/
vi Contents

4 Introduction to Regression Analysis 41

4.1 Preview 41
4.2 Association versus Causality 42
4.3 Statistical versus Deterministic Models 45
4.4 Concluding Remarks 45
References 46

5 Straight-line Regression Analysis 47

5.1 Preview 47
5.2 Regression with a Single Independent Variable 47
5.3 Mathematical Properties of a Straight line 50
5.4 Statistical Assumptions for a Straight-line Model 51
5.5 Determining the Best-fitting Straight Line 55
5.6 Measure of the Quality of the Straight-line Fit and
Estimate of s2 60
5.7 Inferences about the Slope and Intercept 61
5.8 Interpretations of Tests for Slope and Intercept 64
5.9 The Mean Value of Y at a Specified Value of X 66
5.10 Prediction of a New Value of Y at X0 68
5.11 Assessing the Appropriateness of the Straight-line Model 70
5.12 Example: BRFSS Analysis 71
Problems 74
References 107

6 The Correlation Coefficient and Straight-line

Regression Analysis 108
6.1 Definition of r 108
6.2 r as a Measure of Association 109
6.3 The Bivariate Normal Distribution 112
6.4 r 2 and the Strength of the Straight-line Relationship 113
6.5 What r 2 Does Not Measure 115
6.6 Tests of Hypotheses and Confidence Intervals for the
Correlation Coefficient 117
6.7 Testing for the Equality of Two Correlations 120
6.8 Example: BRFSS Analysis 122
6.9 How Large Should r 2 Be in Practice? 123
Problems 125
References 127

7 The Analysis-of-Variance Table 129

7.1 Preview 129
7.2 The ANOVA Table for Straight-line Regression 129
Problems 133

8 Multiple Regression Analysis:

General Considerations 136
8.1 Preview 136
8.2 Multiple Regression Models 137
8.3 Graphical Look at the Problem 138
8.4 Assumptions of Multiple Regression 141
8.5 Determining the Best Estimate of the Multiple
Regression Equation 143
8.6 The ANOVA Table for Multiple Regression 145
8.7 Example: BRFSS Analysis 146
8.8 Numerical Examples 148
Problems 151
References 164

9 Statistical Inference in Multiple Regression 165

9.1 Preview 165
9.2 Test for Significant Overall Regression 166
9.3 Partial F Test 167
9.4 Multiple Partial F Test 172
9.5 Strategies for Using Partial F Tests 175
9.6 Additional Inference Methods for Multiple Regression 180
9.7 Example: BRFSS Analysis 186
Problems 188
References 198

10 Correlations: Multiple, Partial, and Multiple Partial 199

10.1 Preview 199
10.2 Correlation Matrix 200
10.3 Multiple Correlation Coefficient 201
10.4 Relationship of RY | X1, X2, p , XK to the Multivariate
Normal Distribution 203

10.5 Partial Correlation Coefficient 204

10.6 Alternative Representation of the Regression Model 212
10.7 Multiple Partial Correlation 212
10.8 Concluding Remarks 214
Problems 215
References 225

11 Confounding and Interaction in Regression 226

11.1 Preview 226
11.2 Overview 226
11.3 Interaction in Regression 228
11.4 Confounding in Regression 236
11.5 Summary and Conclusions 242
Problems 242
References 256

12 Dummy Variables in Regression 257

12.1 Preview 257
12.2 Definitions 257
12.3 Rule for Defining Dummy Variables 258
12.4 Comparing Two Straight-line Regression Equations: An Example 259
12.5 Questions for Comparing Two Straight Lines 261
12.6 Methods of Comparing Two Straight Lines 262
12.7 Method I: Using Separate Regression Fits to Compare Two Straight Lines 263
12.8 Method II: Using a Single Regression Equation to Compare
Two Straight Lines 268
12.9 Comparison of Methods I and II 271
12.10 Testing Strategies and Interpretation: Comparing Two Straight Lines 272
12.11 Other Dummy Variable Models 273
12.12 Comparing Four Regression Equations 275
12.13 Comparing Several Regression Equations Involving Two Nominal Variables 277
Problems 283
References 307

13 Analysis of Covariance and Other Methods

for Adjusting Continuous Data 308
13.1 Preview 308
13.2 Adjustment Problem 309

13.3 Analysis of Covariance 310

13.4 Assumption of Parallelism: A Potential Drawback 313
13.5 Analysis of Covariance: Several Groups and Several Covariates 314
13.6 Analysis of Covariance: Several Nominal Independent Variables 316
13.7 Comments and Cautions 318
13.8 Summary 321
Problems 321
References 338

14 Regression Diagnostics 339

14.1 Preview 339
14.2 Simple Approaches to Diagnosing Problems in Data 340
14.3 Residual Analysis: Detecting Outliers and Violations of Model Assumptions 347
14.4 Strategies for Addressing Violations of Regression Assumptions 355
14.5 Collinearity 358
14.6 Diagnostics Example 372
Problems 382
References 399

15 Polynomial Regression 401

15.1 Preview 401
15.2 Polynomial Models 402
15.3 Least-squares Procedure for Fitting a Parabola 402
15.4 ANOVA Table for Second-order Polynomial Regression 404
15.5 Inferences Associated with Second-order Polynomial Regression 405
15.6 Example Requiring a Second-order Model 406
15.7 Fitting and Testing Higher-order Models 410
15.8 Lack-of-fit Tests 410
15.9 Orthogonal Polynomials 412
15.10 Strategies for Choosing a Polynomial Model 422
Problems 423

16 Selecting the Best Regression Equation 438

16.1 Preview 438
16.2 Steps in Selecting the Best Regression Equation: Prediction Goal 439
16.3 Step 1: Specifying the Maximum Model: Prediction Goal 439
16.4 Step 2: Specifying a Criterion for Selecting a Model: Prediction Goal 442
16.5 Step 3: Specifying a Strategy for Selecting Variables: Prediction Goal 444
16.6 Step 4: Conducting the Analysis: Prediction Goal 454

16.7 Step 5: Evaluating Reliability with Split Samples: Prediction Goal 454
16.8 Example Analysis of Actual Data 457
16.9 Selecting the Most Valid Model 463
Problems 466
References 480

17 One-way Analysis of Variance 481

17.1 Preview 481
17.2 One-way ANOVA: The Problem, Assumptions, and
Data Configuration 484
17.3 Methodology for One-way Fixed-effects ANOVA 488
17.4 Regression Model for Fixed-effects One-way ANOVA 494
17.5 Fixed-effects Model for One-way ANOVA 497
17.6 Random-effects Model for One-way ANOVA 500
17.7 Multiple-comparison Procedures for Fixed-effects One-way ANOVA 503
17.8 Choosing a Multiple-comparison Technique 515
17.9 Orthogonal Contrasts and Partitioning an ANOVA Sum of Squares 516
Problems 522
References 543

18 Randomized Blocks: Special Case

of Two-way ANOVA 545
18.1 Preview 545
18.2 Equivalent Analysis of a Matched-pairs Experiment 549
18.3 Principle of Blocking 553
18.4 Analysis of a Randomized-blocks Study 555
18.5 ANOVA Table for a Randomized-blocks Study 557
18.6 Regression Models for a Randomized-blocks Study 561
18.7 Fixed-effects ANOVA Model for a Randomized-blocks Study 565
Problems 566
References 578

19 Two-way ANOVA with Equal Cell Numbers 579

19.1 Preview 579
19.2 Using a Table of Cell Means 581
19.3 General Methodology 586
19.4 F Tests for Two-way ANOVA 592
19.5 Regression Model for Fixed-effects Two-way ANOVA 594

19.6 Interactions in Two-way ANOVA 599

19.7 Random- and Mixed-effects Two-way ANOVA Models 607
Problems 610
References 629

20 Two-way ANOVA with Unequal Cell Numbers 630

20.1 Preview 630
20.2 Presentation of Data for Two-way ANOVA: Unequal Cell Numbers 630
20.3 Problem with Unequal Cell Numbers: Nonorthogonality 632
20.4 Regression Approach for Unequal Cell Sample Sizes 637
20.5 Higher-way ANOVA 641
Problems 642
References 659

21 The Method of Maximum Likelihood 661

21.1 Preview 661
21.2 The Principle of Maximum Likelihood 661
21.3 Statistical Inference Using Maximum Likelihood 665
21.4 Summary 677
Problems 678
References 680

22 Logistic Regression Analysis 681

22.1 Preview 681
22.2 The Logistic Model 681
22.3 Estimating the Odds Ratio Using Logistic Regression 683
22.4 A Numerical Example of Logistic Regression 689
22.5 Theoretical Considerations 698
22.6 An Example of Conditional ML Estimation Involving Pair-matched
Data with Unmatched Covariates 704
22.7 Summary 707
Problems 708
References 712

23 Polytomous and Ordinal Logistic Regression 714

23.1 Preview 714
23.2 Why Not Use Binary Regression? 715

23.3 An Example of Polytomous Logistic Regression: One Predictor,

Three Outcome Categories 715
23.4 An Example: Extending the Polytomous Logistic Model to Several Predictors 721
23.5 Ordinal Logistic Regression: Overview 726
23.6 A “Simple” Example: Three Ordinal Categories and
One Dichotomous Exposure Variable 727
23.7 Ordinal Logistic Regression Example Using Real Data with
Four Ordinal Categories and Three Predictor Variables 731
23.8 Summary 737
Problems 738
References 742

24 Poisson Regression Analysis 743

24.1 Preview 743
24.2 The Poisson Distribution 743
24.3 An Example of Poisson Regression 745
24.4 Poisson Regression 748
24.5 Measures of Goodness of Fit 753
24.6 Continuation of Skin Cancer Data Example 756
24.7 A Second Illustration of Poisson Regression Analysis 762
24.8 Summary 765
Problems 766
References 780

25 Analysis of Correlated Data Part 1:

The General Linear Mixed Model 781
25.1 Preview 781
25.2 Examples 784
25.3 General Linear Mixed Model Approach 792
25.4 Example: Study of Effects of an Air Pollution Episode on FEV1 Levels 806
25.5 Summary—Analysis of Correlated Data: Part 1 818
Problems 819
References 824

26 Analysis of Correlated Data Part 2: Random Effects

and Other Issues 825
26.1 Preview 825
26.2 Random Effects Revisited 825

26.3 Results for Models with Random Effects Applied to

Air Pollution Study Data 829
26.4 Second Example—Analysis of Posture Measurement Data 839
26.5 Recommendations about Choice of Correlation Structure 859
26.6 Analysis of Data for Discrete Outcomes 861
Problems 862
References 882

27 Sample Size Planning for Linear and Logistic

Regression and Analysis of Variance 883
27.1 Preview 883
27.2 Review: Sample Size Calculations for Comparisons of
Means and Proportions 884
27.3 Sample Size Planning for Linear Regression 886
27.4 Sample Size Planning for Logistic Regression 889
27.5 Power and Sample Size Determination for Linear Models:
A General Approach 893
27.6 Sample Size Determination for Matched Case–control Studies with a
Dichotomous Outcome 908
27.7 Practical Considerations and Cautions 910
Problems 911
References 913

Appendix A—Tables 915

A.1Standard Normal Cumulative Probabilities 916
A.2Percentiles of the t Distribution 919
A.3Percentiles of the Chi-square Distribution 920
A.4Percentiles of the F Distribution 921
11r
A.5 Values of 12 ln 928
12r
A.6 Upper a Point of Studentized Range 930
A.7 Orthogonal Polynomial Coefficients 932
A.8A Bonferroni Corrected Jackknife Residual Critical Values 933
A.8B Bonferroni Corrected Studentized Residual Critical Values 933
A.9 Critical Values for Leverages 934
A.10 Critical Values for the Maximum of n Values of Cook’s (n 2 k 2 1) di 936

Appendix B—Matrices and Their Relationship

to Regression Analysis 937

Appendix C—SAS Computer Appendix 949

Appendix D—Answers to Selected Problems 991

Index 1037

Copyright 2013 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
Preface
This is the fourth revision of our second-level statistics text, originally published in 1978 and
revised in 1987, 1998, and 2008. As with previous versions, this text is intended primarily
for advanced undergraduates, graduate students, and working professionals in the health,
social, biological, and behavioral sciences who engage in applied research in their fields. The
text may also provide professional statisticians with some new insights into the application of
advanced statistical techniques to realistic research problems.
We have attempted in this revision to retain the basic structure and flavor of the earlier
editions, while at the same time making changes to keep pace with current analytic practices
and computer usage in applied research. Notable changes in this fifth edition, discussed in
more detail later, include
i. Clarification of content and/or terminology as suggested by reviewers and read-
ers, including revision of variable and subscript notation used for predictor vari-
ables and regression coefficients to provide consistency over different chapters.
ii. Expanded and updated coverage of some content areas (e.g., confounding and
interaction in regression in Chapter 11, selecting the best regression equation in
Chapter 16, sample size determination in Chapter 27).
iii. A new linear regression example that is carried through and expanded upon in
Chapters 5, 6, 8, 9, 11, 12, 13, and 16.
iv. Some new exercises at the end of selected chapters, including exercises related to
the new example described in item (iii) above.
v. Updated SAS computer output using SAS 9.3 that reflects improvements in out-
put styling.
vi. Two computer appendices on programming procedures for multiple linear regres-
sion models, logistic regression models, Poisson regression models, and mixed
linear models:
a. In-text: SAS
b. Online: SPSS, STATA, and R
In this fifth edition, as in our previous versions, we emphasize the intuitive logic and
assumptions that underlie the techniques covered, the purposes for which these techniques
are designed, the advantages and disadvantages of these techniques, and valid interpretations
xv
Copyright 2013 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
xvi Preface

based on these techniques. Although we describe the statistical calculations required for the
techniques we cover, we rely on computer output to provide the results of such calculations
so the reader can concentrate on how to apply a given technique rather than how to carry
out the calculations. The mathematical formulas that we do present require no more than
simple algebraic manipulations. Proofs are of secondary importance and are generally omit-
ted. Calculus is not explicitly used anywhere in the main text. We introduce matrix notation
to a limited extent in Chapters 25 and 26 because we believe that the use of matrices provides
a more convenient way to understand some of the complicated mathematical aspects of the
analysis of correlated data. We also have continued to include an appendix on matrices for
the interested reader.
This edition, as with the previous editions, is not intended to be a general reference
work dealing with all the statistical techniques available for analyzing data involving several
variables. Instead, we focus on the techniques we consider most essential for use in applied
research. We want the reader to understand the concepts and assumptions involved in these
techniques and how these techniques can be applied in practice, including how computer
packages can help make it easier to perform the analysis of one’s data.
The most notable features of this fifth edition, including the material that has not been
modified from the previous edition, are the following:

1. Regression analysis (Chapters 1–16) and analysis of variance (Chapters 17–20)

are discussed in considerable detail and with pedagogical care that reflects the
authors’ extensive experience and insight as teachers of such material.
2. A new linear regression example based on a complex survey design is carried
through and expanded upon in several chapters, including new exercises involv-
ing the dataset for this example. To obtain the most valid estimates of regression
coefficients, weighting and stratification schemes involved in the survey design
should be taken into account. Although it is beyond the scope of this text to
describe regression methods for analyzing complex survey designs, we discuss and
illustrate the extent to which results from using such “weighted” methods may
differ from results from using the “unweighted” methods emphasized in this text.
3. The relationship between regression analysis and analysis of variance is high-
lighted.
4. The connection between multiple regression analysis and multiple and partial
correlation analysis is discussed in detail.
5. Several advanced topics are presented in a unique, nonmathematical manner,
including chapters on maximum likelihood (ML) methods (21), binary logis-
tic regression (22), polytomous and ordinal logistic regression (23), and Poisson
regression (24) and two chapters (25–26) on the analysis of correlated data
(described further below). The material on ML methods in Chapters 21–26 pro-
vides a strong foundation for understanding why ML estimation is the most
widely used method for fitting mathematical models involving several variables.
6. An up-to-date discussion of the issues and procedures involved in fine-tuning
a regression analysis is presented on confounding and interaction in regression
(Chapter 11), selecting the best regression model (Chapter 16), and regression
diagnostics (Chapter 14).

7. Chapter 23 on polytomous and ordinal logistic regression methods extends the

standard (binary) logistic model to outcome variables that have more than two
categories. Polytomous logistic regression is used when the outcome categories
do not have any natural order, whereas ordinal logistic regression is appropriate
when the outcome categories have a natural order.
8. Chapters 25 and 26 on the analysis of correlated data describe the ML/REML
linear mixed model approach incorporated into SAS’s MIXED procedure. Since
ML estimation is assumed, these chapters are logically ordered after the current
Chapter 21 on ML estimation. In Chapter 25, we describe the general form
of the linear mixed model, introduce the terms correlated structure and robust/
empirical standard errors, and illustrate how to model correlated data when only
fixed effects are considered. In Chapter 26, which serves as Part 2 of this topic,
we focus on linear mixed models that contain random effects. Chapter 26 also
provides a link to ANOVA Chapters 17–20, alternatively formulating the linear
mixed model approach in terms of an ANOVA that partitions sources of varia-
tion from various predictors into the sums of squares and corresponding mean
square terms of a summary ANOVA table.
9. Chapter 27 on sample size determination for linear and logistic regression models
describes two approaches for sample size calculation, the first being an approxi-
mate approach that yields fairly accurate sample sizes and requires only manual
computation. The second approach is based on more traditional theory for sam-
ple size determination and is best implemented using computer software. This
chapter has been updated to reflect updated SAS 9.3 and PASS 11 output.
10. Representative computer results from SAS 9.3 are used to illustrate concepts in
the body of the text, as well as to provide a basis for exercises for the reader. In
this edition, we revised the computer output to reflect the most recent version of
SAS, and, in many instances, we annotated comments on the output so that it is
easier to read.
11. Numerous examples and exercises illustrate applications to real studies in a wide
variety of disciplines. New exercises have been added to several chapters.
12. Solutions to selected exercises are provided in Appendix D. An Instructor’s Solu-
tions Manual containing solutions to all exercises is also available with the fifth
edition. In addition, a Student Solutions Manual containing complete solutions
to selected problems is available for students.
13. Computer Appendix C is a new addition to the text that describes how to use
Version 9.3 of SAS to carry out linear regression, logistic regression, Poisson
regression, and correlated data analysis of linear models.
14. Links to freely downloadable datasets; a computer appendix on the use of STATA,
SPSS, and R packages to carry out linear regression modeling; updates on errata;
and other information are available at CengageBrain.com.
15. The computer appendix mentioned in item (14) will be a freely downloadable
electronic document providing computer guidelines for multiple linear regression
models. (Other textbooks by Kleinbaum and Klein have computer appendices
for SAS, STATA, and SPSS use with logistic models and Cox proportional

hazards and extended Cox models for survival data.) The computer appendix
will provide a quick and easy reference guide to help the reader avoid having to
spend a lot of time finding information from sometimes confusing help guides in
packages like SAS.

Suggestions for Instructors or Individual Learners

For formal classroom instruction and/or individual/distance learning, the chapters fall natu-
rally into four clusters:
Course 1: Chapters 4–16, on linear regression analysis
Course 2: Chapters 17–20, on the analysis of variance
Course 3: Chapters 21–-24, on maximum likelihood methods and important appli-
cations involving logistic and Poisson regression modeling
Course 4: Chapters 25–26, on the analysis of correlated data involving linear mixed
models
Portions of Chapter 27 on sample size determination could be added, as appropriate, to
Courses 1–3 above. Courses 1 and 2 have often been combined into one course on regression
and ANOVA methods. For a first course in regression analysis, some of Chapters 11 through
16 may be considered too specialized. For example, Chapter 15 on selecting the best regres-
sion model and Chapter 16 on regression diagnostics might be used in a continuation course
on regression modeling, which might also include some of the advanced topics covered in
Chapters 21–27.

Acknowledgments
We wish to acknowledge several people who contributed to the development of this text,
including early editions as well as this fifth edition. Drs. Kleinbaum and Kupper continue to
be indebted to John Cassel and Bernard Greenberg, two mentors who have provided us with
inspiration and the professional and administrative guidance that enabled us at the begin-
ning of our careers to gain the broad experience necessary to write this text.
Dr. Kleinbaum also wishes to thank John Boring, former Chair of the Department of
Epidemiology at Emory University, for his strong support and encouragement during the
writing of the third and fourth editions and for his deep commitment to teaching excellence.
Dr. Kleinbaum also wishes to thank Dr. Mitch Klein of Emory’s Department of Epidemiology
for his colleagueship, including thoughtful suggestions on and review of previous editions.
Dr. Kleinbaum also thanks Dr. Viola Vaccarino, Chair of the Department of Epidemiology
at Emory University, for continued support and encouragement of his academic life at the
Rollins School of Public Health at Emory University.
Dr. Kupper will forever be indebted to Dr. William Mendenhall, founder and longtime
Chair of the University of Florida Department of Statistics. Dr. Mendenhall gave Dr. Kupper
his start in the field of statistics, and he served as a perfect example of an inspiring teacher
and a caring mentor.
Mr. Nizam wishes to thank Dr. Lance Waller, Chair of the Department of Biostatistics
and Bioinformatics at Emory University, for his strong support and Dr. John Spurrier of the
Department of Statistics at the University of South Carolina for being a wonderful teacher,
advisor, and mentor.

We thank Julia Labadie for her assistance in preparing SAS computer output for this
edition. We also thank Dr. Keith Muller for his contributions to earlier editions as one of
our coauthors.
We thank our spouses—Edna Kleinbaum, Sandy Martin, Janet Nizam, and Abby
Horowitz—for their encouragement and support during the writing of various revisions.
We thank our reviewers of the fifth edition for their helpful suggestions:
Joseph Glaz, University of Connecticut
Lynn Kuo, University of Connecticut
Robert Paige, Missouri University of Science and Technology
Debaraj Sen, Concordia University
Po Yang, DePaul University
We thank the Cengage Learning Statistics and Mathematics team, especially Molly
Taylor, Senior Product Manager, and Laura Wheel, Senior Content Developer, for guiding
us through the publication process for the fifth edition, as well as Jessica Rasile, Content
Project Manager, and Tania Andrabi, Production Manager.
David G. Kleinbaum
Lawrence L. Kupper
Azhar Nizam
Eli S. Rosenberg

1.1 Concepts
The purpose of most empirical research is to assess relationships among a set of vari-
ables, which are factors that are distinctly measured on observational units (or subjects).
Multivariable1 techniques are concerned with the statistical analysis of such relationships,
particularly when at least three variables are involved. Regression analysis, our primary focus,
is one type of multivariable technique. Other techniques will also be described in this text.
Choosing an appropriate technique depends on the purpose of the research and on the types
of variables under investigation (a subject discussed in Chapter 2).
Research may be classified broadly into three types: experimental, quasi-experimental, or
observational. Multivariable techniques are applicable to all such types, yet the confidence
one may reasonably have in the results of a study can vary with the research type. In most
types, one variable is usually taken to be a response or dependent variable—that is, a variable
to be predicted from other variables. The other variables are called predictor or independent
variables.
If observational units (subjects) are randomly assigned to levels of important predictors,
the study is usually classified as an experiment. Experiments are the most controlled type of
study; they maximize the investigator’s ability to isolate the observed effect of the predictors
from the distorting effects of other (independent) variables that might also be related to the
response.

1
The term multivariable is preferable to multivariate. Statisticians generally use the term multivariate analysis to
describe a method in which several dependent variables can be considered simultaneously. Researchers in the bio-
medical and health sciences who are not statisticians, however, use this term to describe any statistical technique
involving several variables, even if only one dependent variable is considered at a time. In this text, we prefer to avoid
the confusion by using the term multivariable analysis to denote the latter, more general description.

1
Copyright 2013 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
2 Chapter 1 Concepts and Examples of Research

If subjects are assigned to treatment conditions without randomization, the study is

called quasi-experimental (Campbell and Stanley 1963). Such studies are often more feasible
and less expensive than experimental studies, but they provide less control over the study
situation.
Finally, if all observations are obtained without either randomization or artificial manip-
ulation (i.e., allocation) of the predictor variables, the study is said to be observational. Exper-
iments offer the greatest potential for drawing definitive conclusions, and observational
studies the least; however, experiments are the most difficult studies to implement, and
observational studies the easiest. A researcher must consider this trade-off between interpre-
tive potential and complexity of design when choosing among types of studies (Kleinbaum,
Kupper, and Morgenstern 1982, Chapter 3).
To assess a relationship between two variables, one must measure both of them in some
manner. Measurement inherently and unavoidably involves error. The need for statisti-
cal design and analysis emanates from the presence of such error. Traditionally, statistical
inference has been divided into two kinds: estimation and hypothesis testing. Estimation
refers to describing (i.e., quantifying) characteristics and strengths of relationships. Testing
refers to specifying hypotheses about relationships, making statements of probability about
the reasonableness of such hypotheses, and then providing practical conclusions based on
such statements.
This text focuses on regression and correlation methods involving one response variable
and one or more predictor variables. In these methods, a mathematical model is specified
that describes how the variables of interest are related to one another. The model must some-
how be developed from study data, after which inference-making procedures (e.g., testing
hypotheses and constructing confidence intervals) are conducted about important param-
eters of interest. Although other multivariable regression methods will be discussed, linear
regression techniques are emphasized for three reasons: they have wide applicability; they can
be the most straightforward to implement; and other, more complex statistical procedures
can be better appreciated once linear regression methods are understood.

1.2 Examples
The examples that follow concern real problems from a variety of disciplines and involve
variables to which the methods described in this book can be applied. We shall return to
these examples later when illustrating various methods of multivariable analysis.

■ Example 1.1 Study of the associations among the physician–patient relationship,

perception of pregnancy, and outcome of pregnancy, illustrating the use of regression
analysis and logistic regression analysis.
Thompson (1972) and Hulka and others (1971) looked at both the process and the
outcomes of medical care in a cohort of 107 pregnant married women in North Carolina.
The data were obtained through patient interviews, questionnaires completed by physicians,
and a review of medical records. Several variables were recorded for each patient.
One research goal of primary interest was to determine what association, if any, existed
between satisfaction with medical care and a number of variables meant to describe patient

perception of pregnancy and the physician–patient relationship. Three perception-of-

pregnancy variables measured the patient’s worry during pregnancy, her desire for the baby,
and her concern about childbirth. Two other variables measured the physician–patient rela-
tionship in terms of informational communication concerning prescriptions and affective
communication concerning perceptions. Other variables considered were age, social class,
education, and parity.
Regression analysis was used to describe the relationship between scores measuring patient
satisfaction with medical care and the preceding variables. From this analysis, variables found
not to be related to medical care satisfaction could be eliminated, while those found to be
associated with satisfaction could be ranked in order of importance. Also, the effects of con-
founding variables such as age and social class could be considered, to three ends: any asso-
ciations found could not be attributed solely to such variables; measures of the strength of
the relationship between satisfaction and other variables could be obtained; and a functional
equation predicting level of patient satisfaction in terms of the other variables found to be
important in describing satisfaction could be developed.
Another question of interest in this study was whether patient perception of pregnancy
and/or the physician–patient relationship was associated with complications of pregnancy.
A variable describing complications was defined so that the value 1 could be assigned if the
patient experienced one or more complications of pregnancy and 0 if she experienced no
complications. Logistic regression analysis was used to evaluate the relationship between the
occurrence of complications of pregnancy and other variables. This method, like regression
analysis, allows the researcher to determine and rank important variables that can distinguish
between patients who have complications and patients who do not. ■

■ Example 1.2 Study of race and social influence in cooperative problem-solving dyads,
illustrating the use of analysis of variance and analysis of covariance.
James (1973) conducted an experiment on 140 seventh- and eighth-grade males to
investigate the effects of two factors—race of the experimenter (E) and race of the compari-
son norm (N)—on social influence behaviors in three types of dyads: white–white; black–
black; and white–black. Subjects played a game of strategy called Kill the Bull, in which
14 separate decisions must be made for proceeding toward a defined goal on a game board.
In the game, each pair of players (dyad) must reach a consensus on a direction at each deci-
sion step, after which they signal the E, who then rolls a die to determine how far they can
advance along their chosen path of six squares. Photographs of the current champion players
(N) (either two black youths [black norm] or two white youths [white norm]) were placed
above the game board.
Four measures of social influence activity were used as the outcome variables of inter-
est. One of these, called performance output, was a measure of the number of times a given
subject attempted to influence his dyad to move in a particular direction.
The major research question focused on the outcomes for biracial dyads. Previous
research of this type had used only white investigators and implicit white comparison
norms, and the results indicated that the white partner tended to dominate the decision
making. James’s study sought to determine whether such an “interaction disability,” previ-
ously attributed to blacks, would be maintained, removed, or reversed when the comparison
norm, the experimenter, or both were black. One approach to analyzing this problem was to

perform a two-way analysis of variance on social-influence-activity difference scores between

black and white partners, to assess whether such differences were affected by either the race
of E or the race of N. No such significant effects were found, however, implying that nei-
ther E nor N influenced biracial dyad interaction. Nevertheless, through use of analysis of
covariance, it was shown that, controlling for factors such as age, height, grade, and verbal
and mathematical test scores, there was no statistical evidence of white dominance in any of
the experimental conditions.
Furthermore, when combined output scores for both subjects in same-race dyads
(white–white or black–black) were analyzed using a three-way analysis of variance (the three
factors being race of dyad, race of E, and race of N), subjects in all-black dyads were found
to be more verbally active (i.e., exhibited a greater tendency to influence decisions) under a
black E than under a white E; the same result was found for white dyads under a white E.
This property is generally referred to in statistical jargon as a “race of dyad” by “race of E”
interaction. The property continued to hold up after analysis of covariance was used to con-
trol for the effects of age, height, and verbal and mathematical test scores. ■

■ Example 1.3 Study of the relationship of cultural change to health, illustrating the use
of analysis of variance.
Patrick and others (1974) studied the effects of cultural change on health in the U.S.
Trust Territory island of Ponape. Medical and sociological data were obtained on a sample
of about 2,000 people by means of physical exams and a sociological questionnaire. This
Micronesian island has experienced rapid Westernization and modernization since American
occupation in 1945. The question of primary interest was whether rapid social and cultural
change caused increases in blood pressure and in the incidence of coronary heart disease. A
specific hypothesis guiding the research was that persons with high levels of cultural ambigu-
ity and incongruity and low levels of supportive affiliations with others have high levels of
blood pressure and are at high risk for coronary heart disease.
A preliminary step in the evaluation of this hypothesis involved measuring three vari-
ables: attitude toward modern life; preparation for modern life; and involvement in modern
life. Each of these variables was created by isolating specific questions from a sociological
questionnaire. Then a factor analysis2 determined how best to combine the scores on spe-
cific questions into a single overall score that defined the variable under consideration. Two
cultural incongruity variables were then defined. One involved the discrepancy between
attitude toward modern life and involvement in modern life; the other was defined as the
discrepancy between preparation for modern life and involvement in modern life.
These variables were then analyzed to determine their relationship, if any, to blood pres-
sure and coronary heart disease. Individuals with large positive or negative scores on either
of the two incongruity variables were hypothesized to have high blood pressure and to be at
high risk for coronary heart disease.
One approach to analysis involved categorizing both discrepancy scores into high and
low groups. Then a two-way analysis of variance could be performed using blood pressure

2
Factor analysis was described in Chapter 24 of the second edition of this text, but this topic is not included as a topic
in this (fifth) edition.

as the outcome variable. We will see later that this problem can also be described as a regres-
sion problem. ■

■ Example 1.4 Study of the association between alcohol consumption frequency and
body-mass index (BMI) in the Behavioral Risk Factor Surveillance System (BRFSS).
The BRFSS is a large and ongoing surveillance project managed by the U.S. Centers
for Disease Control and Prevention (CDC) and conducted by state health departments as
telephone-based interviews, based on random-digit dialing. Its purpose is to “generate infor-
mation about health risk behaviors, clinical preventive practices, and health care access and
use primarily related to chronic diseases and injury”(CDC 2012).
The unpublished example considered here examines the relationship between frequency
of alcohol use in the previous 30 days and the response variable of BMI, a common measure
of body fat defined as (weight in kg)Y(height in m)2. Dozens of studies have demonstrated
cardiovascular benefits of red wine consumption. Yet the relationship between alcohol con-
sumption and BMI, an important risk factor for numerous chronic diseases, is less clear.
An analysis of data from the National Health Interview Survey found a moderate reduction
in BMI associated with increasing drinking frequency, yet an increase in BMI with greater
drinking volume (Breslow and Smothers 2005). These relationships were different for males
and females (an example of interaction; see Chapter 11), who are known to metabolize alco-
hol differently.
This analysis of drinking frequency and BMI considers females who live in the state of
Georgia and who consume nonheavy amounts of alcohol (for the 2010 BRFSS data collec-
tion year). Straight-line regression analysis is used to quantify the same negative association
between drinking frequency and BMI found by others. Multiple regression analysis and analy-
sis of covariance are used to additionally consider the effects of age and other health behaviors
(e.g., sleep quality, exercise, and tobacco use) that are known to be associated with BMI.
This example is unique in that it provides key illustrations of the objectives of regres-
sion techniques for the analysis of public health surveillance data on a health outcome with
numerous determinants. These objectives can differ from those used for the analysis of
data emanating from more controlled health studies (such as randomized controlled clini-
cal trials). In particular, the large sample size associated with the BRFSS provides oppor-
tunities for the detection of statistically significant (and sometimes both unexpected and
meaningful) associations between certain determinants and BMI that might otherwise be
challenging to detect. Such hypothesis-generating regression findings can suggest avenues
for further research. It is important to mention that such surveillance studies limit causal
interpretations of the findings. These and related issues are discussed further in several
chapters that follow.

1.3 Concluding Remarks

The four examples described in Section 1.2 indicate the variety of research questions to
which multivariable statistical methods are applicable. In Chapter 2, we will provide a broad
overview of such techniques; in the remaining chapters, we will discuss each technique in
detail.

References
Breslow, R. A., and Smothers, B. A. 2005. “Drinking Patterns and Body Mass Index in Never Smokers:
National Health Interview Survey, 1997–2001.” American Journal of Epidemiology 161(4):
368–76.
Campbell, D. T., and Stanley, J. C. 1963. Experimental and Quasi-experimental Designs for Research.
Chicago: Rand McNally.
CDC Office of Surveillance, Epidemiology, and Laboratory Services. 2012. “Behavioral Risk Factor
Surveillance System: BRFSS Frequently Asked Questions (FAQs).” http://www.cdc.gov/brfss/
faqs.htm.
Hulka, B. S.; Kupper, L. L.; Cassel, J. C.; and Thompson, S. J. 1971. “A Method for Measuring
Physicians’ Awareness of Patients’ Concerns.” HSMHA Health Reports 86: 741–51.
James, S. A. 1973. “The Effects of the Race of Experimenter and Race of Comparison Norm on Social
Influence in Same Race and Biracial Problem-Solving Dyads.” Ph.D. dissertation, Department
of Clinical Psychology, Washington University, St. Louis, Mo.
Kleinbaum, D. G.; Kupper, L. L.; and Morgenstern, H. 1982. Epidemiologic Research. Belmont, Calif.:
Lifetime Learning Publications.
Patrick, R.; Cassel, J. C.; Tyroler, H. A.; Stanley, L.; and Wild, J. 1974. “The Ponape Study of
Health Effects of Cultural Change.” Paper presented at the annual meeting of the Society for
Epidemiologic Research, Berkeley, Calif.
Thompson, S. J. 1972. “The Doctor–Patient Relationship and Outcomes of Pregnancy.” Ph.D.
dissertation, Department of Epidemiology, University of North Carolina, Chapel Hill.

2.1 Classification of Variables

Variables can be classified in a number of ways. Such classifications are useful for determining
which method of data analysis to use. In this section, we describe three methods of classifica-
tion: by gappiness, by level of measurement, and by descriptive orientation.

2.1.1 Gappiness
In the classification scheme we call gappiness, we determine whether gaps exist between
successively observed values of a variable (Figure 2.1). If gaps exist between observations, the
variable is said to be discrete; if no gaps exist, the variable is said to be continuous. To speak
more precisely, a variable is discrete if, between any two potentially observable values, a value
exists that is not possibly observable. A variable is continuous if, between any two potentially
observable values, another potentially observable value exists.
Examples of continuous variables are age, blood pressure, cholesterol level, height, and
weight. Discrete variables are often counts, such as of the numbers of deaths or car accidents.
Additionally, nonnumeric information is often numerically coded in data sources using dis-
crete variables. Examples of this are sex (e.g., 0 if male and 1 if female), group identification
(e.g., 1 if group A and 2 if group B), and state of disease (e.g., 1 if a coronary heart disease
case and 0 if not a coronary heart disease case).
© Cengage Learning

Gaps No gaps

(a) Values of a discrete variable (b) Values of a continuous variable

FIGURE 2.1 Discrete versus continuous variables

7
Copyright 2013 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s).
Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.
8 Chapter 2 Classification of Variables and the Choice of Analysis

Relative frequency

FIGURE 2.2 Sample frequency distributions of a continuous and a discrete variable

In analyses of actual data, the sampling frequency distributions for continuous variables
are represented differently from those for discrete variables. Data on a continuous variable
are usually grouped into class intervals, and a relative frequency distribution is determined
by counting the proportion of observations in each interval. Such a distribution is usually
represented by a histogram, as shown in Figure 2.2(a). Data on a discrete variable, on the
other hand, are usually not grouped but are represented instead by a line chart, as shown in
Figure 2.2(b).
Discrete variables can sometimes be treated for analysis purposes as continuous variables.
This is possible when the values of such a variable, even though discrete, are not far apart
and cover a wide range of numbers. In such a case, the possible values, although technically
gappy, show such small gaps between values that a visual representation would approximate
an interval (Figure 2.3).
Furthermore, a line chart, like the one in Figure 2.2(b), representing the frequency dis-
tribution of data on such a variable would probably show few frequencies greater than 1 and
thus would be uninformative. As an example, the variable “social class” is usually measured as
discrete; one measure of social class1 takes on integer values between 11 and 77. When data
on this variable are grouped into classes (e.g., 11–15, 16–20, etc.), the resulting frequency
histogram gives a clearer picture of the characteristics of the variable than a line chart does.
Thus, in this case, treating social class as a continuous variable is sometimes more useful than
treating it as discrete.
Just as it is often useful to treat a discrete variable as continuous, some fundamentally
continuous variables may be grouped into categories and treated as discrete variables in a
given analysis. For example, the variable “age” can be made discrete by grouping its values
into two categories, “young” and “old.” Similarly, “blood pressure” becomes a discrete vari-
able if it is categorized into “low,” “medium,” and “high” groups or into deciles.

FIGURE 2.3 Discrete variable that may be treated as continuous (© Cengage Learning)

1
Hollingshead’s “Two-Factor Index of Social Position,” a description of which can be found in Green (1970).

The decision to categorize a continuous variable into discrete levels is nuanced, requiring
consideration of both pros and cons. On the one hand, a discrete version of a variable might
make the data easier to collect and summarize. This often, in turn, aids in the presentation of
results to colleagues. Yet these advantages must be balanced against the loss of information
that comes with converting a continuous variable into a discrete one. The choice of variable
type often impacts the type of analysis that can ultimately be conducted, and the desire to use
a certain analysis technique may drive decisions about the treatment of variables.
A further consideration concerns when to categorize continuous data. One may catego-
rize a continuous variable either at the time of data collection or at the time of data analysis.
The former choice often allows cheaper, quicker, and/or less precise methodology for data
collection to be employed. Yet this may also introduce human error (e.g., when a clini-
cian is given the extra step of classifying a continuous reading into one of several groups).
Categorization at the time of analysis reduces the likelihood of human error and also allows
for multiple classification schemes to be later considered, since the original continuous data
have not been forfeited.
A related issue is that both continuous and discrete variables can be error-prone. Contin-
uous variables can be measured with error, and discrete variables can be misclassified. When
such error-prone variables are used in regression analyses, incorrect statistical conclusions can
be made (i.e., statistical validity can be compromised). In this textbook, it will be assumed
that variables to be considered are not subject to either measurement error or misclassifica-
tion error. A discussion of rigorous statistical methods for dealing with error-prone variables
in regression analyses is beyond the scope of this textbook, but Gustafson (2004) provides
numerous relevant references to such methods.

2.1.2 Level of Measurement

A second classification scheme deals with the preciseness of measurement of the variable.
There are three such levels: nominal, ordinal, and interval.
The numerically weakest level of measurement is the nominal level. At this level, the
values assumed by a variable simply indicate different categories. The variable “sex,” for
example, is nominal: by assigning the numbers 1 and 0 to denote male and female, respec-
tively, we distinguish the two sex categories. A variable that describes treatment group is
also nominal, provided that the treatments involved cannot be ranked according to some
criterion (e.g., dosage level).
A somewhat higher level of measurement allows not only grouping into separate cat-
egories but also ordering of categories. This level is called ordinal. The treatment group may
be considered ordinal if, for example, different treatments differ by dosage. In this case, we
could tell not only which treatment group an individual falls into but also who received a
heavier dose of the treatment. Social class is another ordinal variable, since an ordering can
be made among its different categories. For example, all members of the upper middle class
are higher in some sense than all members of the lower middle class.
A limitation—perhaps debatable—in the preciseness of a measurement such as social
class is the amount of information supplied by the magnitude of the differences between
different categories. Thus, although upper middle class is higher than lower middle class, it
is debatable how much higher.

A variable that can give not only an ordering but also a meaningful measure of the
istance between categories is called an interval variable. To be interval, a variable must be
d
expressed in terms of some standard or well-accepted physical unit of measurement. Height,
weight, blood pressure, and number of deaths all satisfy this requirement, whereas subjective
measures such as perception of pregnancy, personality type, prestige, and social stress do not.
An interval variable that has a scale with a true zero is occasionally designated as a
ratio or ratio-scale variable. An example of a ratio-scale variable is the height of a person.
Temperature is commonly measured in degrees Celsius, an interval scale. Measurement of
temperature in degrees Kelvin is based on a scale that begins at absolute zero and thus is a
ratio variable. An example of a ratio variable common in health studies is the concentration
of a substance (e.g., cholesterol) in the blood.
Ratio-scale variables often involve measurement errors that follow a nonnormal
distribution and are proportional to the size of the measurement. We will see in Chapter 5
that such proportional errors violate an important assumption of linear regression—namely,
equality of error variance for all observations. Hence, the presence of a ratio variable is a
signal to be on guard for a possible violation of this assumption. In Chapter 14 (on regression
diagnostics), we will describe methods for detecting and dealing with this problem.
As with variables in other classification schemes, the same variable may be considered at
one level of measurement in one analysis and at a different level in another analysis. Thus,
“age” may be considered as interval in a regression analysis or, by being grouped into catego-
ries, as nominal in an analysis of variance.
The various levels of mathematical preciseness are cumulative. An ordinal scale possesses
all the properties of a nominal scale plus ordinality. An interval scale is also nominal and
ordinal. The cumulativeness of these levels allows the researcher to drop back one or more lev-
els of measurement in analyzing the data. Thus, an interval variable may be treated as nominal
or ordinal for a particular analysis, and an ordinal variable may be analyzed as nominal.

2.1.3 Descriptive Orientation

A third scheme for classifying variables is based on whether a variable is intended to describe
or be described by other variables. Such a classification depends on the study objectives rather
than on the inherent mathematical structure of the variable itself. If the variable under
investigation is to be described in terms of other variables, we call it a response or dependent
variable, typically denoted by the letter Y. If we are using the variable in conjunction with
other variables to describe a given response variable, we call it a predictor, a regressor, or an
independent variable,2 typically denoted by the letter X. Some independent variables may

2
The term independent variable is a historical term meant to evoke the notion that these measured factors may freely
vary from subject to subject, whereas changes in the dependent variable are thought to depend on and be determined by
the values of a subject’s independent variables. This usage of the term independent differs from the statistical concept
of independence. Two variables are statistically independent when the statistical behavior of one variable is completely
unaffected by the statistical behavior of the other variable. When two variables are independent, they are uncorrelated,
although zero correlation does not imply independence. In most regression analysis situations, there are nonzero cor-
relations among the independent (or predictor) variables. Though not ideal terminology, the phrase independent variable
is still commonly used in practice to denote a predictor variable in regression analysis, and we use this standard termi-
nology in this textbook.

Perception Worry
Perception Desire
Perception Birth
Informational Communication
Affective Communication Satisfaction

6
Social Class
Age Control
Education variables

Independent variables Dependent variable

FIGURE 2.4 Descriptive orientation for Thompson’s (1972) study of satisfaction with
medical care

affect relationships among other independent variables and/or the dependent variables but
be of no intrinsic interest in a particular study. Such variables may be referred to as control or
nuisance variables or, in some contexts, as covariates or confounders.
For example, in Thompson’s (1972) study of the relationship between patient per-
ception of pregnancy and patient satisfaction with medical care, the perception variables
are independent variables (or regressors), and the satisfaction variable is the dependent
(or response) variable (Figure 2.4).
Usually, the distinction between independent and dependent variables is clear, as it is in
the examples we have given. Nevertheless, a variable considered as dependent for purposes
of evaluating one study objective may be considered as independent for purposes of evaluat-
ing a different objective. For example, in Thompson’s study, in addition to determining the
relationship of perceptions as independent variables to patient satisfaction, the researcher
sought to determine the relationships of social class, age, and education to perceptions
treated as dependent variables.

2.2 Overlapping of Classification Schemes

The three classification schemes described in Section 2.1 overlap in the sense that any vari-
able can be labeled according to each scheme. “Social class,” for example, may be considered
as ordinal, discrete, and independent in a given study; “blood pressure” may be considered
interval, continuous, and dependent in the same or another study.
The overlap between the level-of-measurement classification and the gappiness classifi-
cation is shown in Figure 2.5. The diagram does not include classification into dependent or
independent variables because that dimension is entirely a function of the study objectives

Interval

Different

Co
representations

nti
of variable “age” Ordinal

nuo
us
Nominal Variable “sex”

and not of the variable itself. In reading the diagram, one should consider any variable as
being representable by some point within the triangle. If the point falls below the dashed
line within the triangle, it is classified as discrete; if it falls above that line, it is continuous.
Also, a point that falls into the area marked “interval” is classified as an interval variable, and
similarly for the other two levels of measurement.
As Figure 2.5 indicates, any nominal variable must be discrete, but a discrete variable
may be nominal, ordinal, or interval. Also, a continuous variable must be either ordinal
or interval, although ordinal or interval variables may exist that are not continuous. For
example, “sex” is nominal and discrete; “age” may be considered interval and continuous or,
if grouped into categories, nominal and discrete; and “social class,” depending on how it is
measured and on the viewpoint of the researcher, may be considered ordinal and continuous,
ordinal and discrete, or nominal and discrete.

2.3 Choice of Analysis

Any researcher faced with the need to analyze data requires a rationale for choosing a par-
ticular method of analysis. Four considerations should enter into such a choice: the purpose
of the investigation; the mathematical characteristics of the variables involved; the statistical
assumptions made about these variables; and how the data are collected (e.g., the sampling
procedure). The first two considerations are generally sufficient to determine an appropriate
analysis. However, the researcher must consider the latter two items before finalizing initial
recommendations.
Here we focus on the use of variable classification, as it relates to the first two
considerations noted at the beginning of this section, in choosing an appropriate method
of analysis. Table 2.1 provides a rough guide to help the researcher in this choice when sev-
eral variables are involved. The guide distinguishes among various multivariable methods.

Table 2.1 Rough guide to multivariable methods

Classification of Variables

Method Dependent Independent General Purpose

Multiple Continuous Classically all To describe the extent, direction, and strength of the
linear regres- continuous, but relationship between several independent variables and a
sion analysis in practice any continuous dependent variable
type(s) can be
used

Analysis of Continuous All nominal To describe the relationship between a continuous

variance dependent variable and one or more nominal independent
variables

Analysis of Continuous Mixture of To describe the relationship between a continuous

covariance nominal variables dependent variable and one or more nominal indepen-
and continuous dent variables, controlling for the effect of one or more
variables (the latter continuous independent variables
used as control
variables)*

Logistic Dichotomous A mixture of vari- To determine how one or more independent variables are
regression ous types can be related to the probability of the occurrence of one of two
analysis used possible outcomes

Poisson Discrete A mixture of vari- To determine how one or more independent variables
regression ous types can be are related to the rate of occurrence of some outcome
analysis used

*Generally, a control variable is a variable that must be considered before any relationships of interest can be quantified; this is because a
control variable may be related to the variables of primary interest and must be taken into account in studying the relationships among the
primary variables. For example, in describing the relationship between blood pressure and physical activity, we would probably consider “age”
and “sex” as control variables because they are related to blood pressure and physical activity and, unless taken into account, could confound
any conclusions regarding the primary relationship of interest.

It considers the types of variable sets usually associated with each method and gives a gen-
eral description of the purposes of each method. In addition to using the table, however,
one must carefully check the statistical assumptions being made. These assumptions will be
described fully later in the text. Table 2.2 shows how these guidelines can be applied to the
examples given in Chapter 1.
Several methods for dealing with multivariable problems are not included in Table 2.1
or in this text—among them, nonparametric methods of analysis of variance, multivariate
multiple regression, and multivariate analysis of variance (which are extensions of the cor-
responding methods given here that allow for several dependent variables), as well as methods
of cluster analysis. In this book, we will cover only the multivariable techniques used most
often by health and social researchers.

Newala, too, suffers from the distance of its water-supply—at least

the Newala of to-day does; there was once another Newala in a lovely
valley at the foot of the plateau. I visited it and found scarcely a trace
of houses, only a Christian cemetery, with the graves of several
missionaries and their converts, remaining as a monument of its
former glories. But the surroundings are wonderfully beautiful. A
thick grove of splendid mango-trees closes in the weather-worn
crosses and headstones; behind them, combining the useful and the
agreeable, is a whole plantation of lemon-trees covered with ripe
fruit; not the small African kind, but a much larger and also juicier
imported variety, which drops into the hands of the passing traveller,
without calling for any exertion on his part. Old Newala is now under
the jurisdiction of the native pastor, Daudi, at Chingulungulu, who,
as I am on very friendly terms with him, allows me, as a matter of
course, the use of this lemon-grove during my stay at Newala.
FEET MUTILATED BY THE RAVAGES OF THE “JIGGER”
(Sarcopsylla penetrans)

The water-supply of New Newala is in the bottom of the valley,

some 1,600 feet lower down. The way is not only long and fatiguing,
but the water, when we get it, is thoroughly bad. We are suffering not
only from this, but from the fact that the arrangements at Newala are
nothing short of luxurious. We have a separate kitchen—a hut built
against the boma palisade on the right of the baraza, the interior of
which is not visible from our usual position. Our two cooks were not
long in finding this out, and they consequently do—or rather neglect
to do—what they please. In any case they do not seem to be very
particular about the boiling of our drinking-water—at least I can
attribute to no other cause certain attacks of a dysenteric nature,
from which both Knudsen and I have suffered for some time. If a
man like Omari has to be left unwatched for a moment, he is capable
of anything. Besides this complaint, we are inconvenienced by the
state of our nails, which have become as hard as glass, and crack on
the slightest provocation, and I have the additional infliction of
pimples all over me. As if all this were not enough, we have also, for
the last week been waging war against the jigger, who has found his
Eldorado in the hot sand of the Makonde plateau. Our men are seen
all day long—whenever their chronic colds and the dysentery likewise
raging among them permit—occupied in removing this scourge of
Africa from their feet and trying to prevent the disastrous
consequences of its presence. It is quite common to see natives of
this place with one or two toes missing; many have lost all their toes,
or even the whole front part of the foot, so that a well-formed leg
ends in a shapeless stump. These ravages are caused by the female of
Sarcopsylla penetrans, which bores its way under the skin and there
develops an egg-sac the size of a pea. In all books on the subject, it is
stated that one’s attention is called to the presence of this parasite by
an intolerable itching. This agrees very well with my experience, so
far as the softer parts of the sole, the spaces between and under the
toes, and the side of the foot are concerned, but if the creature
penetrates through the harder parts of the heel or ball of the foot, it
may escape even the most careful search till it has reached maturity.
Then there is no time to be lost, if the horrible ulceration, of which
we see cases by the dozen every day, is to be prevented. It is much
easier, by the way, to discover the insect on the white skin of a
European than on that of a native, on which the dark speck scarcely
shows. The four or five jiggers which, in spite of the fact that I
constantly wore high laced boots, chose my feet to settle in, were
taken out for me by the all-accomplished Knudsen, after which I
thought it advisable to wash out the cavities with corrosive
sublimate. The natives have a different sort of disinfectant—they fill
the hole with scraped roots. In a tiny Makua village on the slope of
the plateau south of Newala, we saw an old woman who had filled all
the spaces under her toe-nails with powdered roots by way of
prophylactic treatment. What will be the result, if any, who can say?
The rest of the many trifling ills which trouble our existence are
really more comic than serious. In the absence of anything else to
smoke, Knudsen and I at last opened a box of cigars procured from
the Indian store-keeper at Lindi, and tried them, with the most
distressing results. Whether they contain opium or some other
narcotic, neither of us can say, but after the tenth puff we were both
“off,” three-quarters stupefied and unspeakably wretched. Slowly we
recovered—and what happened next? Half-an-hour later we were
once more smoking these poisonous concoctions—so insatiable is the
craving for tobacco in the tropics.
Even my present attacks of fever scarcely deserve to be taken
seriously. I have had no less than three here at Newala, all of which
have run their course in an incredibly short time. In the early
afternoon, I am busy with my old natives, asking questions and
making notes. The strong midday coffee has stimulated my spirits to
an extraordinary degree, the brain is active and vigorous, and work
progresses rapidly, while a pleasant warmth pervades the whole
body. Suddenly this gives place to a violent chill, forcing me to put on
my overcoat, though it is only half-past three and the afternoon sun
is at its hottest. Now the brain no longer works with such acuteness
and logical precision; more especially does it fail me in trying to
establish the syntax of the difficult Makua language on which I have
ventured, as if I had not enough to do without it. Under the
circumstances it seems advisable to take my temperature, and I do
so, to save trouble, without leaving my seat, and while going on with
my work. On examination, I find it to be 101·48°. My tutors are
abruptly dismissed and my bed set up in the baraza; a few minutes
later I am in it and treating myself internally with hot water and
lemon-juice.
Three hours later, the thermometer marks nearly 104°, and I make
them carry me back into the tent, bed and all, as I am now perspiring
heavily, and exposure to the cold wind just beginning to blow might
mean a fatal chill. I lie still for a little while, and then find, to my
great relief, that the temperature is not rising, but rather falling. This
is about 7.30 p.m. At 8 p.m. I find, to my unbounded astonishment,
that it has fallen below 98·6°, and I feel perfectly well. I read for an
hour or two, and could very well enjoy a smoke, if I had the
wherewithal—Indian cigars being out of the question.
Having no medical training, I am at a loss to account for this state
of things. It is impossible that these transitory attacks of high fever
should be malarial; it seems more probable that they are due to a
kind of sunstroke. On consulting my note-book, I become more and
more inclined to think this is the case, for these attacks regularly
follow extreme fatigue and long exposure to strong sunshine. They at
least have the advantage of being only short interruptions to my
work, as on the following morning I am always quite fresh and fit.
My treasure of a cook is suffering from an enormous hydrocele which
makes it difficult for him to get up, and Moritz is obliged to keep in
the dark on account of his inflamed eyes. Knudsen’s cook, a raw boy
from somewhere in the bush, knows still less of cooking than Omari;
consequently Nils Knudsen himself has been promoted to the vacant
post. Finding that we had come to the end of our supplies, he began
by sending to Chingulungulu for the four sucking-pigs which we had
bought from Matola and temporarily left in his charge; and when
they came up, neatly packed in a large crate, he callously slaughtered
the biggest of them. The first joint we were thoughtless enough to
entrust for roasting to Knudsen’s mshenzi cook, and it was
consequently uneatable; but we made the rest of the animal into a
jelly which we ate with great relish after weeks of underfeeding,
consuming incredible helpings of it at both midday and evening
meals. The only drawback is a certain want of variety in the tinned
vegetables. Dr. Jäger, to whom the Geographical Commission
entrusted the provisioning of the expeditions—mine as well as his
own—because he had more time on his hands than the rest of us,
seems to have laid in a huge stock of Teltow turnips,[46] an article of
food which is all very well for occasional use, but which quickly palls
when set before one every day; and we seem to have no other tins
left. There is no help for it—we must put up with the turnips; but I
am certain that, once I am home again, I shall not touch them for ten
years to come.
Amid all these minor evils, which, after all, go to make up the
genuine flavour of Africa, there is at least one cheering touch:
Knudsen has, with the dexterity of a skilled mechanic, repaired my 9
× 12 cm. camera, at least so far that I can use it with a little care.
How, in the absence of finger-nails, he was able to accomplish such a
ticklish piece of work, having no tool but a clumsy screw-driver for
taking to pieces and putting together again the complicated
mechanism of the instantaneous shutter, is still a mystery to me; but
he did it successfully. The loss of his finger-nails shows him in a light
contrasting curiously enough with the intelligence evinced by the
above operation; though, after all, it is scarcely surprising after his
ten years’ residence in the bush. One day, at Lindi, he had occasion
to wash a dog, which must have been in need of very thorough
cleansing, for the bottle handed to our friend for the purpose had an
extremely strong smell. Having performed his task in the most
conscientious manner, he perceived with some surprise that the dog
did not appear much the better for it, and was further surprised by
finding his own nails ulcerating away in the course of the next few
days. “How was I to know that carbolic acid has to be diluted?” he
mutters indignantly, from time to time, with a troubled gaze at his
mutilated finger-tips.
Since we came to Newala we have been making excursions in all
directions through the surrounding country, in accordance with old
habit, and also because the akida Sefu did not get together the tribal
elders from whom I wanted information so speedily as he had
promised. There is, however, no harm done, as, even if seen only
from the outside, the country and people are interesting enough.
The Makonde plateau is like a large rectangular table rounded off
at the corners. Measured from the Indian Ocean to Newala, it is
about seventy-five miles long, and between the Rovuma and the
Lukuledi it averages fifty miles in breadth, so that its superficial area
is about two-thirds of that of the kingdom of Saxony. The surface,
however, is not level, but uniformly inclined from its south-western
edge to the ocean. From the upper edge, on which Newala lies, the
eye ranges for many miles east and north-east, without encountering
any obstacle, over the Makonde bush. It is a green sea, from which
here and there thick clouds of smoke rise, to show that it, too, is
inhabited by men who carry on their tillage like so many other
primitive peoples, by cutting down and burning the bush, and
manuring with the ashes. Even in the radiant light of a tropical day
such a fire is a grand sight.
Much less effective is the impression produced just now by the
great western plain as seen from the edge of the plateau. As often as
time permits, I stroll along this edge, sometimes in one direction,
sometimes in another, in the hope of finding the air clear enough to
let me enjoy the view; but I have always been disappointed.
Wherever one looks, clouds of smoke rise from the burning bush,
and the air is full of smoke and vapour. It is a pity, for under more
favourable circumstances the panorama of the whole country up to
the distant Majeje hills must be truly magnificent. It is of little use
taking photographs now, and an outline sketch gives a very poor idea
of the scenery. In one of these excursions I went out of my way to
make a personal attempt on the Makonde bush. The present edge of
the plateau is the result of a far-reaching process of destruction
through erosion and denudation. The Makonde strata are
everywhere cut into by ravines, which, though short, are hundreds of
yards in depth. In consequence of the loose stratification of these
beds, not only are the walls of these ravines nearly vertical, but their
upper end is closed by an equally steep escarpment, so that the
western edge of the Makonde plateau is hemmed in by a series of
deep, basin-like valleys. In order to get from one side of such a ravine
to the other, I cut my way through the bush with a dozen of my men.
It was a very open part, with more grass than scrub, but even so the
short stretch of less than two hundred yards was very hard work; at
the end of it the men’s calicoes were in rags and they themselves
bleeding from hundreds of scratches, while even our strong khaki
suits had not escaped scatheless.

NATIVE PATH THROUGH THE MAKONDE BUSH, NEAR

MAHUTA

I see increasing reason to believe that the view formed some time
back as to the origin of the Makonde bush is the correct one. I have
no doubt that it is not a natural product, but the result of human
occupation. Those parts of the high country where man—as a very
slight amount of practice enables the eye to perceive at once—has not
yet penetrated with axe and hoe, are still occupied by a splendid
timber forest quite able to sustain a comparison with our mixed
forests in Germany. But wherever man has once built his hut or tilled
his field, this horrible bush springs up. Every phase of this process
may be seen in the course of a couple of hours’ walk along the main
road. From the bush to right or left, one hears the sound of the axe—
not from one spot only, but from several directions at once. A few
steps further on, we can see what is taking place. The brush has been
cut down and piled up in heaps to the height of a yard or more,
between which the trunks of the large trees stand up like the last
pillars of a magnificent ruined building. These, too, present a
melancholy spectacle: the destructive Makonde have ringed them—
cut a broad strip of bark all round to ensure their dying off—and also
piled up pyramids of brush round them. Father and son, mother and
son-in-law, are chopping away perseveringly in the background—too
busy, almost, to look round at the white stranger, who usually excites
so much interest. If you pass by the same place a week later, the piles
of brushwood have disappeared and a thick layer of ashes has taken
the place of the green forest. The large trees stretch their
smouldering trunks and branches in dumb accusation to heaven—if
they have not already fallen and been more or less reduced to ashes,
perhaps only showing as a white stripe on the dark ground.
This work of destruction is carried out by the Makonde alike on the
virgin forest and on the bush which has sprung up on sites already
cultivated and deserted. In the second case they are saved the trouble
of burning the large trees, these being entirely absent in the
secondary bush.
After burning this piece of forest ground and loosening it with the
hoe, the native sows his corn and plants his vegetables. All over the
country, he goes in for bed-culture, which requires, and, in fact,
receives, the most careful attention. Weeds are nowhere tolerated in
the south of German East Africa. The crops may fail on the plains,
where droughts are frequent, but never on the plateau with its
abundant rains and heavy dews. Its fortunate inhabitants even have
the satisfaction of seeing the proud Wayao and Wamakua working
for them as labourers, driven by hunger to serve where they were
accustomed to rule.
But the light, sandy soil is soon exhausted, and would yield no
harvest the second year if cultivated twice running. This fact has
been familiar to the native for ages; consequently he provides in
time, and, while his crop is growing, prepares the next plot with axe
and firebrand. Next year he plants this with his various crops and
lets the first piece lie fallow. For a short time it remains waste and
desolate; then nature steps in to repair the destruction wrought by
man; a thousand new growths spring out of the exhausted soil, and
even the old stumps put forth fresh shoots. Next year the new growth
is up to one’s knees, and in a few years more it is that terrible,
impenetrable bush, which maintains its position till the black
occupier of the land has made the round of all the available sites and
come back to his starting point.
The Makonde are, body and soul, so to speak, one with this bush.
According to my Yao informants, indeed, their name means nothing
else but “bush people.” Their own tradition says that they have been
settled up here for a very long time, but to my surprise they laid great
stress on an original immigration. Their old homes were in the
south-east, near Mikindani and the mouth of the Rovuma, whence
their peaceful forefathers were driven by the continual raids of the
Sakalavas from Madagascar and the warlike Shirazis[47] of the coast,
to take refuge on the almost inaccessible plateau. I have studied
African ethnology for twenty years, but the fact that changes of
population in this apparently quiet and peaceable corner of the earth
could have been occasioned by outside enterprises taking place on
the high seas, was completely new to me. It is, no doubt, however,
correct.
The charming tribal legend of the Makonde—besides informing us
of other interesting matters—explains why they have to live in the
thickest of the bush and a long way from the edge of the plateau,
instead of making their permanent homes beside the purling brooks
and springs of the low country.
“The place where the tribe originated is Mahuta, on the southern
side of the plateau towards the Rovuma, where of old time there was
nothing but thick bush. Out of this bush came a man who never
washed himself or shaved his head, and who ate and drank but little.
He went out and made a human figure from the wood of a tree
growing in the open country, which he took home to his abode in the
bush and there set it upright. In the night this image came to life and
was a woman. The man and woman went down together to the
Rovuma to wash themselves. Here the woman gave birth to a still-
born child. They left that place and passed over the high land into the
valley of the Mbemkuru, where the woman had another child, which
was also born dead. Then they returned to the high bush country of
Mahuta, where the third child was born, which lived and grew up. In
course of time, the couple had many more children, and called
themselves Wamatanda. These were the ancestral stock of the
Makonde, also called Wamakonde,[48] i.e., aborigines. Their
forefather, the man from the bush, gave his children the command to
bury their dead upright, in memory of the mother of their race who
was cut out of wood and awoke to life when standing upright. He also
warned them against settling in the valleys and near large streams,
for sickness and death dwelt there. They were to make it a rule to
have their huts at least an hour’s walk from the nearest watering-
place; then their children would thrive and escape illness.”
The explanation of the name Makonde given by my informants is
somewhat different from that contained in the above legend, which I
extract from a little book (small, but packed with information), by
Pater Adams, entitled Lindi und sein Hinterland. Otherwise, my
results agree exactly with the statements of the legend. Washing?
Hapana—there is no such thing. Why should they do so? As it is, the
supply of water scarcely suffices for cooking and drinking; other
people do not wash, so why should the Makonde distinguish himself
by such needless eccentricity? As for shaving the head, the short,
woolly crop scarcely needs it,[49] so the second ancestral precept is
likewise easy enough to follow. Beyond this, however, there is
nothing ridiculous in the ancestor’s advice. I have obtained from
various local artists a fairly large number of figures carved in wood,
ranging from fifteen to twenty-three inches in height, and
representing women belonging to the great group of the Mavia,
Makonde, and Matambwe tribes. The carving is remarkably well
done and renders the female type with great accuracy, especially the
keloid ornamentation, to be described later on. As to the object and
meaning of their works the sculptors either could or (more probably)
would tell me nothing, and I was forced to content myself with the
scanty information vouchsafed by one man, who said that the figures
were merely intended to represent the nembo—the artificial
deformations of pelele, ear-discs, and keloids. The legend recorded
by Pater Adams places these figures in a new light. They must surely
be more than mere dolls; and we may even venture to assume that
they are—though the majority of present-day Makonde are probably
unaware of the fact—representations of the tribal ancestress.
The references in the legend to the descent from Mahuta to the
Rovuma, and to a journey across the highlands into the Mbekuru
valley, undoubtedly indicate the previous history of the tribe, the
travels of the ancestral pair typifying the migrations of their
descendants. The descent to the neighbouring Rovuma valley, with
its extraordinary fertility and great abundance of game, is intelligible
at a glance—but the crossing of the Lukuledi depression, the ascent
to the Rondo Plateau and the descent to the Mbemkuru, also lie
within the bounds of probability, for all these districts have exactly
the same character as the extreme south. Now, however, comes a
point of especial interest for our bacteriological age. The primitive
Makonde did not enjoy their lives in the marshy river-valleys.
Disease raged among them, and many died. It was only after they
had returned to their original home near Mahuta, that the health
conditions of these people improved. We are very apt to think of the
African as a stupid person whose ignorance of nature is only equalled
by his fear of it, and who looks on all mishaps as caused by evil
spirits and malignant natural powers. It is much more correct to
assume in this case that the people very early learnt to distinguish
districts infested with malaria from those where it is absent.
This knowledge is crystallized in the
ancestral warning against settling in the
valleys and near the great waters, the
dwelling-places of disease and death. At the
same time, for security against the hostile
Mavia south of the Rovuma, it was enacted
that every settlement must be not less than a
certain distance from the southern edge of the
plateau. Such in fact is their mode of life at the
present day. It is not such a bad one, and
certainly they are both safer and more
comfortable than the Makua, the recent
intruders from the south, who have made USUAL METHOD OF
good their footing on the western edge of the CLOSING HUT-DOOR
plateau, extending over a fairly wide belt of
country. Neither Makua nor Makonde show in their dwellings
anything of the size and comeliness of the Yao houses in the plain,
especially at Masasi, Chingulungulu and Zuza’s. Jumbe Chauro, a
Makonde hamlet not far from Newala, on the road to Mahuta, is the
most important settlement of the tribe I have yet seen, and has fairly
spacious huts. But how slovenly is their construction compared with
the palatial residences of the elephant-hunters living in the plain.
The roofs are still more untidy than in the general run of huts during
the dry season, the walls show here and there the scanty beginnings
or the lamentable remains of the mud plastering, and the interior is a
veritable dog-kennel; dirt, dust and disorder everywhere. A few huts
only show any attempt at division into rooms, and this consists
merely of very roughly-made bamboo partitions. In one point alone
have I noticed any indication of progress—in the method of fastening
the door. Houses all over the south are secured in a simple but
ingenious manner. The door consists of a set of stout pieces of wood
or bamboo, tied with bark-string to two cross-pieces, and moving in
two grooves round one of the door-posts, so as to open inwards. If
the owner wishes to leave home, he takes two logs as thick as a man’s
upper arm and about a yard long. One of these is placed obliquely
against the middle of the door from the inside, so as to form an angle
of from 60° to 75° with the ground. He then places the second piece
horizontally across the first, pressing it downward with all his might.
It is kept in place by two strong posts planted in the ground a few
inches inside the door. This fastening is absolutely safe, but of course
cannot be applied to both doors at once, otherwise how could the
owner leave or enter his house? I have not yet succeeded in finding
out how the back door is fastened.

MAKONDE LOCK AND KEY AT JUMBE CHAURO

This is the general way of closing a house. The Makonde at Jumbe
Chauro, however, have a much more complicated, solid and original
one. Here, too, the door is as already described, except that there is
only one post on the inside, standing by itself about six inches from
one side of the doorway. Opposite this post is a hole in the wall just
large enough to admit a man’s arm. The door is closed inside by a
large wooden bolt passing through a hole in this post and pressing
with its free end against the door. The other end has three holes into
which fit three pegs running in vertical grooves inside the post. The
door is opened with a wooden key about a foot long, somewhat
curved and sloped off at the butt; the other end has three pegs
corresponding to the holes, in the bolt, so that, when it is thrust
through the hole in the wall and inserted into the rectangular
opening in the post, the pegs can be lifted and the bolt drawn out.[50]

MODE OF INSERTING THE KEY

With no small pride first one householder and then a second

showed me on the spot the action of this greatest invention of the
Makonde Highlands. To both with an admiring exclamation of
“Vizuri sana!” (“Very fine!”). I expressed the wish to take back these
marvels with me to Ulaya, to show the Wazungu what clever fellows
the Makonde are. Scarcely five minutes after my return to camp at
Newala, the two men came up sweating under the weight of two
heavy logs which they laid down at my feet, handing over at the same
time the keys of the fallen fortress. Arguing, logically enough, that if
the key was wanted, the lock would be wanted with it, they had taken
their axes and chopped down the posts—as it never occurred to them
to dig them out of the ground and so bring them intact. Thus I have
two badly damaged specimens, and the owners, instead of praise,
come in for a blowing-up.
The Makua huts in the environs of Newala are especially
miserable; their more than slovenly construction reminds one of the
temporary erections of the Makua at Hatia’s, though the people here
have not been concerned in a war. It must therefore be due to
congenital idleness, or else to the absence of a powerful chief. Even
the baraza at Mlipa’s, a short hour’s walk south-east of Newala,
shares in this general neglect. While public buildings in this country
are usually looked after more or less carefully, this is in evident
danger of being blown over by the first strong easterly gale. The only
attractive object in this whole district is the grave of the late chief
Mlipa. I visited it in the morning, while the sun was still trying with
partial success to break through the rolling mists, and the circular
grove of tall euphorbias, which, with a broken pot, is all that marks
the old king’s resting-place, impressed one with a touch of pathos.
Even my very materially-minded carriers seemed to feel something
of the sort, for instead of their usual ribald songs, they chanted
solemnly, as we marched on through the dense green of the Makonde
bush:—
“We shall arrive with the great master; we stand in a row and have
no fear about getting our food and our money from the Serkali (the
Government). We are not afraid; we are going along with the great
master, the lion; we are going down to the coast and back.”
With regard to the characteristic features of the various tribes here
on the western edge of the plateau, I can arrive at no other
conclusion than the one already come to in the plain, viz., that it is
impossible for anyone but a trained anthropologist to assign any
given individual at once to his proper tribe. In fact, I think that even
an anthropological specialist, after the most careful examination,
might find it a difficult task to decide. The whole congeries of peoples
collected in the region bounded on the west by the great Central
African rift, Tanganyika and Nyasa, and on the east by the Indian
Ocean, are closely related to each other—some of their languages are
only distinguished from one another as dialects of the same speech,
and no doubt all the tribes present the same shape of skull and
structure of skeleton. Thus, surely, there can be no very striking
differences in outward appearance.
Even did such exist, I should have no time
to concern myself with them, for day after day,
I have to see or hear, as the case may be—in
any case to grasp and record—an
extraordinary number of ethnographic
phenomena. I am almost disposed to think it
fortunate that some departments of inquiry, at
least, are barred by external circumstances.
Chief among these is the subject of iron-
working. We are apt to think of Africa as a
country where iron ore is everywhere, so to
speak, to be picked up by the roadside, and
where it would be quite surprising if the
inhabitants had not learnt to smelt the
material ready to their hand. In fact, the
knowledge of this art ranges all over the
continent, from the Kabyles in the north to the
Kafirs in the south. Here between the Rovuma
and the Lukuledi the conditions are not so
favourable. According to the statements of the
Makonde, neither ironstone nor any other
form of iron ore is known to them. They have
not therefore advanced to the art of smelting
the metal, but have hitherto bought all their
THE ANCESTRESS OF
THE MAKONDE
iron implements from neighbouring tribes.
Even in the plain the inhabitants are not much
better off. Only one man now living is said to
understand the art of smelting iron. This old fundi lives close to
Huwe, that isolated, steep-sided block of granite which rises out of
the green solitude between Masasi and Chingulungulu, and whose
jagged and splintered top meets the traveller’s eye everywhere. While
still at Masasi I wished to see this man at work, but was told that,
frightened by the rising, he had retired across the Rovuma, though
he would soon return. All subsequent inquiries as to whether the
fundi had come back met with the genuine African answer, “Bado”
(“Not yet”).
BRAZIER

Some consolation was afforded me by a brassfounder, whom I

came across in the bush near Akundonde’s. This man is the favourite
of women, and therefore no doubt of the gods; he welds the glittering
brass rods purchased at the coast into those massive, heavy rings
which, on the wrists and ankles of the local fair ones, continually give
me fresh food for admiration. Like every decent master-craftsman he
had all his tools with him, consisting of a pair of bellows, three
crucibles and a hammer—nothing more, apparently. He was quite
willing to show his skill, and in a twinkling had fixed his bellows on
the ground. They are simply two goat-skins, taken off whole, the four
legs being closed by knots, while the upper opening, intended to
admit the air, is kept stretched by two pieces of wood. At the lower
end of the skin a smaller opening is left into which a wooden tube is
stuck. The fundi has quickly borrowed a heap of wood-embers from
the nearest hut; he then fixes the free ends of the two tubes into an
earthen pipe, and clamps them to the ground by means of a bent
piece of wood. Now he fills one of his small clay crucibles, the dross
on which shows that they have been long in use, with the yellow
material, places it in the midst of the embers, which, at present are
only faintly glimmering, and begins his work. In quick alternation
the smith’s two hands move up and down with the open ends of the
bellows; as he raises his hand he holds the slit wide open, so as to let
the air enter the skin bag unhindered. In pressing it down he closes
the bag, and the air puffs through the bamboo tube and clay pipe into
the fire, which quickly burns up. The smith, however, does not keep
on with this work, but beckons to another man, who relieves him at
the bellows, while he takes some more tools out of a large skin pouch
carried on his back. I look on in wonder as, with a smooth round
stick about the thickness of a finger, he bores a few vertical holes into
the clean sand of the soil. This should not be difficult, yet the man
seems to be taking great pains over it. Then he fastens down to the
ground, with a couple of wooden clamps, a neat little trough made by
splitting a joint of bamboo in half, so that the ends are closed by the
two knots. At last the yellow metal has attained the right consistency,
and the fundi lifts the crucible from the fire by means of two sticks
split at the end to serve as tongs. A short swift turn to the left—a
tilting of the crucible—and the molten brass, hissing and giving forth
clouds of smoke, flows first into the bamboo mould and then into the
holes in the ground.
The technique of this backwoods craftsman may not be very far
advanced, but it cannot be denied that he knows how to obtain an
adequate result by the simplest means. The ladies of highest rank in
this country—that is to say, those who can afford it, wear two kinds
of these massive brass rings, one cylindrical, the other semicircular
in section. The latter are cast in the most ingenious way in the
bamboo mould, the former in the circular hole in the sand. It is quite
a simple matter for the fundi to fit these bars to the limbs of his fair
customers; with a few light strokes of his hammer he bends the
pliable brass round arm or ankle without further inconvenience to
the wearer.
SHAPING THE POT

SMOOTHING WITH MAIZE-COB

CUTTING THE EDGE

FINISHING THE BOTTOM

LAST SMOOTHING BEFORE

BURNING

FIRING THE BRUSH-PILE

LIGHTING THE FARTHER SIDE OF
THE PILE

TURNING THE RED-HOT VESSEL

NYASA WOMAN MAKING POTS AT MASASI

Pottery is an art which must always and everywhere excite the
interest of the student, just because it is so intimately connected with
the development of human culture, and because its relics are one of
the principal factors in the reconstruction of our own condition in
prehistoric times. I shall always remember with pleasure the two or
three afternoons at Masasi when Salim Matola’s mother, a slightly-
built, graceful, pleasant-looking woman, explained to me with
touching patience, by means of concrete illustrations, the ceramic art
of her people. The only implements for this primitive process were a
lump of clay in her left hand, and in the right a calabash containing
the following valuables: the fragment of a maize-cob stripped of all
its grains, a smooth, oval pebble, about the size of a pigeon’s egg, a
few chips of gourd-shell, a bamboo splinter about the length of one’s
hand, a small shell, and a bunch of some herb resembling spinach.
Nothing more. The woman scraped with the
shell a round, shallow hole in the soft, fine
sand of the soil, and, when an active young
girl had filled the calabash with water for her,
she began to knead the clay. As if by magic it
gradually assumed the shape of a rough but
already well-shaped vessel, which only wanted
a little touching up with the instruments
before mentioned. I looked out with the
MAKUA WOMAN closest attention for any indication of the use
MAKING A POT. of the potter’s wheel, in however rudimentary
SHOWS THE a form, but no—hapana (there is none). The
BEGINNINGS OF THE embryo pot stood firmly in its little
POTTER’S WHEEL
depression, and the woman walked round it in
a stooping posture, whether she was removing
small stones or similar foreign bodies with the maize-cob, smoothing
the inner or outer surface with the splinter of bamboo, or later, after
letting it dry for a day, pricking in the ornamentation with a pointed
bit of gourd-shell, or working out the bottom, or cutting the edge
with a sharp bamboo knife, or giving the last touches to the finished
vessel. This occupation of the women is infinitely toilsome, but it is
without doubt an accurate reproduction of the process in use among
our ancestors of the Neolithic and Bronze ages.
There is no doubt that the invention of pottery, an item in human
progress whose importance cannot be over-estimated, is due to
women. Rough, coarse and unfeeling, the men of the horde range
over the countryside. When the united cunning of the hunters has
succeeded in killing the game; not one of them thinks of carrying
home the spoil. A bright fire, kindled by a vigorous wielding of the
drill, is crackling beside them; the animal has been cleaned and cut
up secundum artem, and, after a slight singeing, will soon disappear
under their sharp teeth; no one all this time giving a single thought
to wife or child.
To what shifts, on the other hand, the primitive wife, and still more
the primitive mother, was put! Not even prehistoric stomachs could
endure an unvarying diet of raw food. Something or other suggested
the beneficial effect of hot water on the majority of approved but
indigestible dishes. Perhaps a neighbour had tried holding the hard
roots or tubers over the fire in a calabash filled with water—or maybe
an ostrich-egg-shell, or a hastily improvised vessel of bark. They
became much softer and more palatable than they had previously
been; but, unfortunately, the vessel could not stand the fire and got
charred on the outside. That can be remedied, thought our
ancestress, and plastered a layer of wet clay round a similar vessel.
This is an improvement; the cooking utensil remains uninjured, but
the heat of the fire has shrunk it, so that it is loose in its shell. The
next step is to detach it, so, with a firm grip and a jerk, shell and
kernel are separated, and pottery is invented. Perhaps, however, the
discovery which led to an intelligent use of the burnt-clay shell, was
made in a slightly different way. Ostrich-eggs and calabashes are not
to be found in every part of the world, but everywhere mankind has
arrived at the art of making baskets out of pliant materials, such as
bark, bast, strips of palm-leaf, supple twigs, etc. Our inventor has no
water-tight vessel provided by nature. “Never mind, let us line the
basket with clay.” This answers the purpose, but alas! the basket gets
burnt over the blazing fire, the woman watches the process of
cooking with increasing uneasiness, fearing a leak, but no leak
appears. The food, done to a turn, is eaten with peculiar relish; and
the cooking-vessel is examined, half in curiosity, half in satisfaction
at the result. The plastic clay is now hard as stone, and at the same
time looks exceedingly well, for the neat plaiting of the burnt basket
is traced all over it in a pretty pattern. Thus, simultaneously with
pottery, its ornamentation was invented.
Primitive woman has another claim to respect. It was the man,
roving abroad, who invented the art of producing fire at will, but the
woman, unable to imitate him in this, has been a Vestal from the
earliest times. Nothing gives so much trouble as the keeping alight of
the smouldering brand, and, above all, when all the men are absent
from the camp. Heavy rain-clouds gather, already the first large
drops are falling, the first gusts of the storm rage over the plain. The
little flame, a greater anxiety to the woman than her own children,
flickers unsteadily in the blast. What is to be done? A sudden thought
occurs to her, and in an instant she has constructed a primitive hut
out of strips of bark, to protect the flame against rain and wind.
This, or something very like it, was the way in which the principle
of the house was discovered; and even the most hardened misogynist
cannot fairly refuse a woman the credit of it. The protection of the
hearth-fire from the weather is the germ from which the human
dwelling was evolved. Men had little, if any share, in this forward
step, and that only at a late stage. Even at the present day, the
plastering of the housewall with clay and the manufacture of pottery
are exclusively the women’s business. These are two very significant
survivals. Our European kitchen-garden, too, is originally a woman’s
invention, and the hoe, the primitive instrument of agriculture, is,
characteristically enough, still used in this department. But the
noblest achievement which we owe to the other sex is unquestionably
the art of cookery. Roasting alone—the oldest process—is one for
which men took the hint (a very obvious one) from nature. It must
have been suggested by the scorched carcase of some animal
overtaken by the destructive forest-fires. But boiling—the process of
improving organic substances by the help of water heated to boiling-
point—is a much later discovery. It is so recent that it has not even
yet penetrated to all parts of the world. The Polynesians understand
how to steam food, that is, to cook it, neatly wrapped in leaves, in a
hole in the earth between hot stones, the air being excluded, and
(sometimes) a few drops of water sprinkled on the stones; but they
do not understand boiling.
To come back from this digression, we find that the slender Nyasa
woman has, after once more carefully examining the finished pot,
put it aside in the shade to dry. On the following day she sends me
word by her son, Salim Matola, who is always on hand, that she is
going to do the burning, and, on coming out of my house, I find her
already hard at work. She has spread on the ground a layer of very
dry sticks, about as thick as one’s thumb, has laid the pot (now of a
yellowish-grey colour) on them, and is piling brushwood round it.
My faithful Pesa mbili, the mnyampara, who has been standing by,
most obligingly, with a lighted stick, now hands it to her. Both of
them, blowing steadily, light the pile on the lee side, and, when the
flame begins to catch, on the weather side also. Soon the whole is in a
blaze, but the dry fuel is quickly consumed and the fire dies down, so
that we see the red-hot vessel rising from the ashes. The woman
turns it continually with a long stick, sometimes one way and
sometimes another, so that it may be evenly heated all over. In
twenty minutes she rolls it out of the ash-heap, takes up the bundle
of spinach, which has been lying for two days in a jar of water, and
sprinkles the red-hot clay with it. The places where the drops fall are
marked by black spots on the uniform reddish-brown surface. With a
sigh of relief, and with visible satisfaction, the woman rises to an
erect position; she is standing just in a line between me and the fire,
from which a cloud of smoke is just rising: I press the ball of my
camera, the shutter clicks—the apotheosis is achieved! Like a
priestess, representative of her inventive sex, the graceful woman
stands: at her feet the hearth-fire she has given us beside her the
invention she has devised for us, in the background the home she has
built for us.
At Newala, also, I have had the manufacture of pottery carried on
in my presence. Technically the process is better than that already
described, for here we find the beginnings of the potter’s wheel,
which does not seem to exist in the plains; at least I have seen
nothing of the sort. The artist, a frightfully stupid Makua woman, did
not make a depression in the ground to receive the pot she was about
to shape, but used instead a large potsherd. Otherwise, she went to
work in much the same way as Salim’s mother, except that she saved
herself the trouble of walking round and round her work by squatting
at her ease and letting the pot and potsherd rotate round her; this is
surely the first step towards a machine. But it does not follow that
the pot was improved by the process. It is true that it was beautifully
rounded and presented a very creditable appearance when finished,
but the numerous large and small vessels which I have seen, and, in
part, collected, in the “less advanced” districts, are no less so. We
moderns imagine that instruments of precision are necessary to
produce excellent results. Go to the prehistoric collections of our
museums and look at the pots, urns and bowls of our ancestors in the
dim ages of the past, and you will at once perceive your error.
MAKING LONGITUDINAL CUT IN
BARK

DRAWING THE BARK OFF THE LOG

REMOVING THE OUTER BARK

BEATING THE BARK

WORKING THE BARK-CLOTH AFTER BEATING, TO MAKE IT

SOFT

MANUFACTURE OF BARK-CLOTH AT NEWALA

To-day, nearly the whole population of German East Africa is
clothed in imported calico. This was not always the case; even now in
some parts of the north dressed skins are still the prevailing wear,
and in the north-western districts—east and north of Lake
Tanganyika—lies a zone where bark-cloth has not yet been
superseded. Probably not many generations have passed since such
bark fabrics and kilts of skins were the only clothing even in the
south. Even to-day, large quantities of this bright-red or drab
material are still to be found; but if we wish to see it, we must look in
the granaries and on the drying stages inside the native huts, where
it serves less ambitious uses as wrappings for those seeds and fruits
which require to be packed with special care. The salt produced at
Masasi, too, is packed for transport to a distance in large sheets of
bark-cloth. Wherever I found it in any degree possible, I studied the
process of making this cloth. The native requisitioned for the
purpose arrived, carrying a log between two and three yards long and
as thick as his thigh, and nothing else except a curiously-shaped
mallet and the usual long, sharp and pointed knife which all men and
boys wear in a belt at their backs without a sheath—horribile dictu!
[51]
Silently he squats down before me, and with two rapid cuts has
drawn a couple of circles round the log some two yards apart, and
slits the bark lengthwise between them with the point of his knife.
With evident care, he then scrapes off the outer rind all round the
log, so that in a quarter of an hour the inner red layer of the bark
shows up brightly-coloured between the two untouched ends. With
some trouble and much caution, he now loosens the bark at one end,
and opens the cylinder. He then stands up, takes hold of the free
edge with both hands, and turning it inside out, slowly but steadily
pulls it off in one piece. Now comes the troublesome work of
scraping all superfluous particles of outer bark from the outside of
the long, narrow piece of material, while the inner side is carefully
scrutinised for defective spots. At last it is ready for beating. Having
signalled to a friend, who immediately places a bowl of water beside
him, the artificer damps his sheet of bark all over, seizes his mallet,
lays one end of the stuff on the smoothest spot of the log, and
hammers away slowly but continuously. “Very simple!” I think to
myself. “Why, I could do that, too!”—but I am forced to change my
opinions a little later on; for the beating is quite an art, if the fabric is
not to be beaten to pieces. To prevent the breaking of the fibres, the
stuff is several times folded across, so as to interpose several
thicknesses between the mallet and the block. At last the required
state is reached, and the fundi seizes the sheet, still folded, by both
ends, and wrings it out, or calls an assistant to take one end while he
holds the other. The cloth produced in this way is not nearly so fine
and uniform in texture as the famous Uganda bark-cloth, but it is
quite soft, and, above all, cheap.
Now, too, I examine the mallet. My craftsman has been using the
simpler but better form of this implement, a conical block of some
hard wood, its base—the striking surface—being scored across and
across with more or less deeply-cut grooves, and the handle stuck
into a hole in the middle. The other and earlier form of mallet is
shaped in the same way, but the head is fastened by an ingenious
network of bark strips into the split bamboo serving as a handle. The
observation so often made, that ancient customs persist longest in
connection with religious ceremonies and in the life of children, here
finds confirmation. As we shall soon see, bark-cloth is still worn
during the unyago,[52] having been prepared with special solemn
ceremonies; and many a mother, if she has no other garment handy,
will still put her little one into a kilt of bark-cloth, which, after all,
looks better, besides being more in keeping with its African
surroundings, than the ridiculous bit of print from Ulaya.
MAKUA WOMEN

Applied Regression Analysis.
25% (4)
Applied Regression Analysis.
9 pages
Kleinbaum Applied Regression Analysis and Other Multivariable Methods 3 Ed PDF
0% (6)
Kleinbaum Applied Regression Analysis and Other Multivariable Methods 3 Ed PDF
9 pages
Selling Building Partnerships 10Th Edition Ebook PDF All Chapter Scribd Ebook PDF
100% (36)
Selling Building Partnerships 10Th Edition Ebook PDF All Chapter Scribd Ebook PDF
61 pages
Ebook PDF Nutrition Now 8Th Edition Full Chapter
100% (31)
Ebook PDF Nutrition Now 8Th Edition Full Chapter
61 pages
EBook Strengths Based Therapy Connecting Theory Practice and Skills 1St Edition Ebook PDF Version PDF Docx Kindle Full Chapter
100% (35)
EBook Strengths Based Therapy Connecting Theory Practice and Skills 1St Edition Ebook PDF Version PDF Docx Kindle Full Chapter
62 pages
EBOOK Interactions Collaboration Skills For School Professionals 7Th Edition Download Full Chapter PDF Kindle
100% (53)
EBOOK Interactions Collaboration Skills For School Professionals 7Th Edition Download Full Chapter PDF Kindle
61 pages
EBOOK Accounting Fundamentals For Health Care Management 3Rd Edition Ebook PDF Download Full Chapter PDF Kindle
100% (49)
EBOOK Accounting Fundamentals For Health Care Management 3Rd Edition Ebook PDF Download Full Chapter PDF Kindle
62 pages
EBOOK Public Personnel Management Current Concerns Future Challenges 6Th Edition Ebook PDF Version Download Full Chapter PDF Kindle
98% (62)
EBOOK Public Personnel Management Current Concerns Future Challenges 6Th Edition Ebook PDF Version Download Full Chapter PDF Kindle
58 pages
EBOOK Lasers and Lights E Book Procedures in Cosmetic Dermatology Series 4Th Edition Ebook PDF Download Full Chapter PDF Kindle
100% (55)
EBOOK Lasers and Lights E Book Procedures in Cosmetic Dermatology Series 4Th Edition Ebook PDF Download Full Chapter PDF Kindle
58 pages
EBOOK When Words Collide 9Th Edition Ebook PDF Download Full Chapter PDF Kindle
98% (51)
EBOOK When Words Collide 9Th Edition Ebook PDF Download Full Chapter PDF Kindle
52 pages
EBook Nursing The Philosophy and Science of Caring Revised Edition Ebook PDF Version PDF Docx Kindle Full Chapter
100% (39)
EBook Nursing The Philosophy and Science of Caring Revised Edition Ebook PDF Version PDF Docx Kindle Full Chapter
58 pages
EBOOK System For Ophthalmic Dispensing E Book 3Rd Edition Ebook PDF Download Full Chapter PDF Kindle
100% (48)
EBOOK System For Ophthalmic Dispensing E Book 3Rd Edition Ebook PDF Download Full Chapter PDF Kindle
61 pages
EBOOK The Social Work Experience A Case Based Introduction To and Welfare 7Th Download Full Chapter PDF Kindle
100% (51)
EBOOK The Social Work Experience A Case Based Introduction To and Welfare 7Th Download Full Chapter PDF Kindle
62 pages
Etextbook PDF For Global Climate Change Turning Knowledge Into Action All Chapter Scribd Ebook PDF
100% (31)
Etextbook PDF For Global Climate Change Turning Knowledge Into Action All Chapter Scribd Ebook PDF
61 pages
EBook Etextbook PDF For Reflect and Relate An Introduction To Interpersonal Communication 5Th Edition PDF Docx Kindle Full Chapter
100% (43)
EBook Etextbook PDF For Reflect and Relate An Introduction To Interpersonal Communication 5Th Edition PDF Docx Kindle Full Chapter
61 pages
EBook Gender Law and Policy Aspen College Series 2Nd Edition Ebook PDF PDF Docx Kindle Full Chapter
100% (38)
EBook Gender Law and Policy Aspen College Series 2Nd Edition Ebook PDF PDF Docx Kindle Full Chapter
61 pages
Ebook Ebook PDF Go All in One Go For Office 2016 Series 3Rd Edition 2 All Chapter PDF Docx Kindle
100% (37)
Ebook Ebook PDF Go All in One Go For Office 2016 Series 3Rd Edition 2 All Chapter PDF Docx Kindle
41 pages
EBook Etextbook 978 1429219617 Raven Biology of Plants 8Th Edition by Ray F Evert PDF Docx Kindle Full Chapter
100% (46)
EBook Etextbook 978 1429219617 Raven Biology of Plants 8Th Edition by Ray F Evert PDF Docx Kindle Full Chapter
62 pages
EBOOK Hospitality Law Managing Legal Issues in The Hospitality Industry 5Th Edition Ebook PDF Version Download Full Chapter PDF Kindle
100% (53)
EBOOK Hospitality Law Managing Legal Issues in The Hospitality Industry 5Th Edition Ebook PDF Version Download Full Chapter PDF Kindle
58 pages
EBOOK Ebook PDF Clinical Cases in Eye Care by Mark Rosenfield Download Full Chapter PDF Kindle
100% (53)
EBOOK Ebook PDF Clinical Cases in Eye Care by Mark Rosenfield Download Full Chapter PDF Kindle
61 pages
Ebook PDF Etextbook PDF For Research Methods From Theory To Practice by Ben Gorvine Full Chapter
100% (42)
Ebook PDF Etextbook PDF For Research Methods From Theory To Practice by Ben Gorvine Full Chapter
62 pages
(eTextbook PDF) for Applied Regression Analysis and Other Multivariable Methods 5th Editionpdf download
100% (2)
(eTextbook PDF) for Applied Regression Analysis and Other Multivariable Methods 5th Editionpdf download
56 pages
(eBook PDF) Applied Regression Analysis and Other Multivariable Methods 5th Edition instant download
100% (2)
(eBook PDF) Applied Regression Analysis and Other Multivariable Methods 5th Edition instant download
46 pages
Download ebooks file (eBook PDF) Applied Regression Analysis and Other Multivariable Methods 5th Edition all chapters
100% (2)
Download ebooks file (eBook PDF) Applied Regression Analysis and Other Multivariable Methods 5th Edition all chapters
36 pages
Instant Download (eTextbook PDF) for Applied Regression Analysis and Other Multivariable Methods 5th Edition PDF All Chapters
100% (11)
Instant Download (eTextbook PDF) for Applied Regression Analysis and Other Multivariable Methods 5th Edition PDF All Chapters
65 pages
High Yield Notes
No ratings yet
High Yield Notes
251 pages
01 - Quantitative Methods
No ratings yet
01 - Quantitative Methods
28 pages
Eco 6
No ratings yet
Eco 6
96 pages
Bio2 Module 4 - Multiple Linear Regression
No ratings yet
Bio2 Module 4 - Multiple Linear Regression
20 pages
Week 5 Multiple Regression: Busa3500 Statistics For Business Ii Piedmont College
No ratings yet
Week 5 Multiple Regression: Busa3500 Statistics For Business Ii Piedmont College
57 pages
Linear Regression
100% (2)
Linear Regression
28 pages
Lecture 25 - Multiple Regression
No ratings yet
Lecture 25 - Multiple Regression
34 pages
Intermediate Analytics-Regression-Week 1
No ratings yet
Intermediate Analytics-Regression-Week 1
52 pages
Business Forecasting J. Holton (1) - 251-300
No ratings yet
Business Forecasting J. Holton (1) - 251-300
50 pages
Correlation
No ratings yet
Correlation
5 pages
Ch08 Part 2 - Multtiple Regression
No ratings yet
Ch08 Part 2 - Multtiple Regression
45 pages
Chapter 3 MLR
No ratings yet
Chapter 3 MLR
40 pages
Unit 3
No ratings yet
Unit 3
24 pages
Chapter 6
No ratings yet
Chapter 6
20 pages
Ch08 Part 2 - Multiple Regression
No ratings yet
Ch08 Part 2 - Multiple Regression
45 pages
11bda
No ratings yet
11bda
25 pages
A Second Course in Statistics Regression Analysis
No ratings yet
A Second Course in Statistics Regression Analysis
8 pages
4.1 Multiple Regression Models
No ratings yet
4.1 Multiple Regression Models
6 pages
A Second Course in Statistics: Regression Analysis: Journal of The American Statistical Association June 1997
No ratings yet
A Second Course in Statistics: Regression Analysis: Journal of The American Statistical Association June 1997
9 pages
Investigating Variables
No ratings yet
Investigating Variables
15 pages
What Is Multiple Linear Regression
No ratings yet
What Is Multiple Linear Regression
23 pages
Dr. Hussin Abdullah School of Economics, Finance and Banking, Uum Cob
No ratings yet
Dr. Hussin Abdullah School of Economics, Finance and Banking, Uum Cob
12 pages
Untitled 472
No ratings yet
Untitled 472
13 pages
Regression
No ratings yet
Regression
24 pages
Regression
No ratings yet
Regression
15 pages
Lecture 12
No ratings yet
Lecture 12
47 pages
Chapter05DemandEstimation (1)
No ratings yet
Chapter05DemandEstimation (1)
41 pages
(eBook PDF) A Second Course in Statistics: Regression Analysis 8th Editioninstant download
100% (3)
(eBook PDF) A Second Course in Statistics: Regression Analysis 8th Editioninstant download
48 pages
Regression analysis
No ratings yet
Regression analysis
16 pages
Interpreting Correlation
No ratings yet
Interpreting Correlation
13 pages
1.1 Regression Analysis
No ratings yet
1.1 Regression Analysis
33 pages
STAT630Slide Adv Data Analysis
No ratings yet
STAT630Slide Adv Data Analysis
238 pages
Chapter 14 Mr
No ratings yet
Chapter 14 Mr
35 pages
Chapter 14 Multiple Regression and Correlation Analysis
No ratings yet
Chapter 14 Multiple Regression and Correlation Analysis
25 pages
Profit From Your Forecasting Software: A Best Practice Guide for Sales Forecasters
From Everand
Profit From Your Forecasting Software: A Best Practice Guide for Sales Forecasters
Paul Goodwin
No ratings yet
Etextbook PDF For America A Narrative History Eleventh Edition Vol Volume 1 11Th Edition All Chapter Scribd Ebook PDF
100% (36)
Etextbook PDF For America A Narrative History Eleventh Edition Vol Volume 1 11Th Edition All Chapter Scribd Ebook PDF
61 pages
Etextbook PDF For Applied Calculus 7Th Edition by Stefan Waner All Chapter Scribd Ebook PDF
100% (34)
Etextbook PDF For Applied Calculus 7Th Edition by Stefan Waner All Chapter Scribd Ebook PDF
61 pages
EBOOK Etextbook PDF For African American Music An Introduction 2Nd Edition Download Full Chapter PDF Docx Kindle
100% (59)
EBOOK Etextbook PDF For African American Music An Introduction 2Nd Edition Download Full Chapter PDF Docx Kindle
61 pages
Etextbook PDF For Administration of Programs For Young Children 9Th Edition All Chapter Scribd Ebook PDF
100% (33)
Etextbook PDF For Administration of Programs For Young Children 9Th Edition All Chapter Scribd Ebook PDF
61 pages

EBook Etextbook PDF For Applied Regression Analysis and Other Multivariable Methods 5Th Edition PDF Docx Kindle Full Chapter

Uploaded by

EBook Etextbook PDF For Applied Regression Analysis and Other Multivariable Methods 5Th Edition PDF Docx Kindle Full Chapter

Uploaded by

(eTextbook PDF) for Applied

Regression Analysis and Other

4 Introduction to Regression Analysis 41

5 Straight-line Regression Analysis 47

6 The Correlation Coefficient and Straight-line

7 The Analysis-of-Variance Table 129

8 Multiple Regression Analysis:

9 Statistical Inference in Multiple Regression 165

10 Correlations: Multiple, Partial, and Multiple Partial 199

10.5 Partial Correlation Coefficient 204

11 Confounding and Interaction in Regression 226

12 Dummy Variables in Regression 257

13 Analysis of Covariance and Other Methods

13.3 Analysis of Covariance 310

14 Regression Diagnostics 339

15 Polynomial Regression 401

16 Selecting the Best Regression Equation 438

17 One-way Analysis of Variance 481

18 Randomized Blocks: Special Case

19 Two-way ANOVA with Equal Cell Numbers 579

19.6 Interactions in Two-way ANOVA 599

20 Two-way ANOVA with Unequal Cell Numbers 630

21 The Method of Maximum Likelihood 661

22 Logistic Regression Analysis 681

23 Polytomous and Ordinal Logistic Regression 714

23.3 An Example of Polytomous Logistic Regression: One Predictor,

24 Poisson Regression Analysis 743

25 Analysis of Correlated Data Part 1:

26 Analysis of Correlated Data Part 2: Random Effects

26.3 Results for Models with Random Effects Applied to

27 Sample Size Planning for Linear and Logistic

Appendix A—Tables 915

Appendix B—Matrices and Their Relationship

Appendix C—SAS Computer Appendix 949

Appendix D—Answers to Selected Problems 991

1. Regression analysis (Chapters 1–16) and analysis of variance (Chapters 17–20)

7. Chapter 23 on polytomous and ordinal logistic regression methods extends the

Suggestions for Instructors or Individual Learners

If subjects are assigned to treatment conditions without randomization, the study is

■ Example 1.1 Study of the associations among the physician–patient relationship,

perception of pregnancy and the physician–patient relationship. Three perception-of-

perform a two-way analysis of variance on social-influence-activity difference scores between

1.3 Concluding Remarks

2.1 Classification of Variables

(a) Values of a discrete variable (b) Values of a continuous variable

FIGURE 2.1 Discrete versus continuous variables

FIGURE 2.2 Sample frequency distributions of a continuous and a discrete variable

2.1.2 Level of Measurement

2.1.3 Descriptive Orientation

Independent variables Dependent variable

2.2 Overlapping of Classification Schemes

2.3 Choice of Analysis

Table 2.1 Rough guide to multivariable methods

Method Dependent Independent General Purpose

Analysis of Continuous All nominal To describe the relationship between a continuous

Analysis of Continuous Mixture of To describe the relationship between a continuous

Newala, too, suffers from the distance of its water-supply—at least

The water-supply of New Newala is in the bottom of the valley,

NATIVE PATH THROUGH THE MAKONDE BUSH, NEAR

MAKONDE LOCK AND KEY AT JUMBE CHAURO

MODE OF INSERTING THE KEY

With no small pride first one householder and then a second

Some consolation was afforded me by a brassfounder, whom I

SMOOTHING WITH MAIZE-COB

CUTTING THE EDGE

LAST SMOOTHING BEFORE

FIRING THE BRUSH-PILE

TURNING THE RED-HOT VESSEL

NYASA WOMAN MAKING POTS AT MASASI

DRAWING THE BARK OFF THE LOG

REMOVING THE OUTER BARK

WORKING THE BARK-CLOTH AFTER BEATING, TO MAKE IT

MANUFACTURE OF BARK-CLOTH AT NEWALA

You might also like

6 The Correlation Coefficient and Straight-line

8 Multiple Regression Analysis:

13 Analysis of Covariance and Other Methods

18 Randomized Blocks: Special Case

23.3 An Example of Polytomous Logistic Regression: One Predictor,

25 Analysis of Correlated Data Part 1:

26 Analysis of Correlated Data Part 2: Random Effects

26.3 Results for Models with Random Effects Applied to

27 Sample Size Planning for Linear and Logistic