stat.js

$.extend(KhanUtil, {
    sum: function(values) {
        var sum = 0;

        $.each(values, function(index, value) {
            sum += value;
        });
        return sum;
    },

    mean: function(values) {
        return KhanUtil.sum(values) / values.length;
    },

    median: function(values) {
        var sortedInts, median;
        sortedInts = KhanUtil.sortNumbers(values);

        if (values.length % 2 === 0) {
            median = KhanUtil.roundTo(1,
                (sortedInts[(values.length / 2) - 1] + sortedInts[values.length / 2]) / 2);
        } else {
            median = sortedInts[Math.floor(values.length / 2)];
        }
        return median;
    },

    mode: function(values) {
        var numInstances = [];
        var modeInstances = -1;

        var mode;
        for (var i = 0; i < values.length; i++) {
            if (!numInstances[values[i]]) {
                numInstances[values[i]] = 1;
            } else {
                numInstances[values[i]] += 1;
                if (numInstances[values[i]] > modeInstances) {
                    modeInstances = numInstances[values[i]];
                    mode = values[i];
                }
            }
        }

        // iterate again to check for 'no mode'
        for (var i = 0; i < numInstances.length; i++) {
            if (numInstances[i]) {
                if (i !== mode && numInstances[i] >= modeInstances) {
                    return false;
                }
            }
        }

        return mode;
    },

    variance: function(values) {
        var xbar = KhanUtil.mean(values);
        var n = values.length;

        var sum = 0;
        $.each(values, function(i, x_i) {
            sum += (x_i - xbar) * (x_i - xbar);
        });
        return sum / (n - 1);
    },

    variancePop: function(values) {
        var xbar = KhanUtil.mean(values);
        var N = values.length;

        var sum = 0;
        $.each(values, function(i, x_i) {
            sum += (x_i - xbar) * (x_i - xbar);
        });
        return sum / N;
    },

    stdDev: function(values) {
        return Math.sqrt(KhanUtil.variance(values));
    },

    stdDevPop: function(values) {
        return Math.sqrt(KhanUtil.variancePop(values));
    },

    // Gaussian distribution using Box-Muller transform
    // defaults to standard normal unless target mean and stddev are passed
    // Pass "count" to get an array of data
    randGaussian: function(tgtMean, tgtStdDev, count) {
        if (count == null) {
            var x1, x2, rad, y1;

            do {
                x1 = 2 * KhanUtil.random() - 1;
                x2 = 2 * KhanUtil.random() - 1;
                rad = x1 * x1 + x2 * x2;
            } while (rad >= 1 || rad == 0);

            var c = Math.sqrt(-2 * Math.log(rad) / rad);

            return x1 * c * (tgtStdDev || 1) + (tgtMean || 0);
        } else {
            return $.map(new Array(count), function() {
                return KhanUtil.randGaussian(tgtMean, tgtStdDev);
            });
        }
    },

    gaussianPDF: function(mean, stddev, x) {
        return (1 / Math.sqrt(2 * Math.PI * stddev * stddev)) * Math.exp(-((x - mean) * (x - mean)) / (2 * stddev * stddev));
    }

});