statsmodels.stats.multicomp.pairwise_tukeyhsd¶

statsmodels.stats.multicomp.pairwise_tukeyhsd(endog, groups, alpha=0.05, use_var='equal')[source]¶

Calculate all pairwise comparisons with TukeyHSD or Games-Howell.

Parameters:¶

endogndarray, float, 1d

response variable

groupsndarray, 1d

array with groups, can be string or integers

alphafloat

significance level for the test

use_var{“unequal”, “equal”}

If use_var is “equal”, then the Tukey-hsd pvalues are returned. Tukey-hsd assumes that (within) variances are the same across groups. If use_var is “unequal”, then the Games-Howell pvalues are returned. This uses Welch’s t-test for unequal variances with Satterthwaite’s corrected degrees of freedom for each pairwise comparison.

Returns:¶

resultsTukeyHSDResults instance

A results class containing relevant data and some post-hoc calculations, including adjusted p-value

See also

MultiComparison

tukeyhsd

statsmodels.sandbox.stats.multicomp.TukeyHSDResults

Notes

This is just a wrapper around tukeyhsd method of MultiComparison. Tukey-hsd is not robust to heteroscedasticity, i.e. variance differ across groups, especially if group sizes also vary. In those cases, the actual size (rejection rate under the Null hypothesis) might be far from the nominal size of the test. The Games-Howell method uses pairwise t-tests that are robust to differences in variances and approximately maintains size unless samples are very small.

Added in version 0.15: The use_var keyword and option for Games-Howell test.