Categories &

Functions List

Function Reference: ansaribradley

statistics: h = ansaribradley (x, y)
statistics: h = ansaribradley (x, y, name, value)
statistics: [h, p] = ansaribradley (…)
statistics: [h, p, stats] = ansaribradley (…)

Ansari-Bradley two-sample test for equal dispersions.

h = ansaribradley (x, y) performs an Ansari-Bradley test of the hypothesis that the two independent samples in the vectors x and y come from distributions with the same dispersion parameter, against the alternative that they come from distributions with different dispersions. The result is h = 0 if the null hypothesis of equal dispersions cannot be rejected at the 5% significance level, or h = 1 if it can.

The Ansari-Bradley test is a nonparametric alternative to the two-sample F test (vartest2) that does not assume normality. It assumes that the two samples are independent and that they come from distributions with the same median and shape, differing (under the alternative) only in dispersion. If the medians differ, the data should be recentred (e.g. by subtracting the sample medians) before applying the test.

ansaribradley treats NaNs in x or y as missing values and ignores them.

[h, p] = ansaribradley (…) returns the p-value of the test, that is the probability, under the null hypothesis, of observing a value of the test statistic as or more extreme than the one observed.

[h, p, stats] = ansaribradley (…) returns a structure with the following fields:

Wthe value of the Ansari-Bradley test statistic, the sum of the Ansari-Bradley scores of the sample x
Wstarthe value of the approximate normal (z) statistic

Source Code: ansaribradley

[…] = ansaribradley (…, name, value) specifies one or more of the following name/value pairs:

NameValue
'alpha'the significance level. Default is 0.05.
'tail'a string specifying the alternative hypothesis
'method'a string selecting the p-value computation, either 'exact' to use the exact permutation distribution of the statistic, or 'approximate' to use the normal approximation. The default is 'exact' when the total sample size is 25 or less, and 'approximate' otherwise.

Source Code: ansaribradley

The 'tail' option can take one of the following values:

'both'dispersions are not equal (two-tailed, default)
'right'dispersion of x is greater than dispersion of y (right-tailed)
'left'dispersion of x is less than dispersion of y (left-tailed)

Source Code: ansaribradley

See also: vartest2, vartestn, kstest2, ranksum

Source Code: ansaribradley

Test whether two samples have the same dispersion. The second sample is drawn with twice the standard deviation, so the null hypothesis of equal dispersions should be rejected.

 x = [42, 44, 38, 52, 48, 46, 40, 50];
 y = [30, 62, 25, 70, 33, 58, 20, 65];
 [h, p, stats] = ansaribradley (x, y)
h = 1
p = 1.5540e-04
stats =

  scalar structure containing the fields:

    W = 52
    Wstar = 3.3806