statsmodels.stats.weightstats._tstat_generic¶

statsmodels.stats.weightstats._tstat_generic(value1, value2, std_diff, dof, alternative, diff=0)[source]¶

generic ttest based on summary statistic

The test statistic is :

tstat = (value1 - value2 - diff) / std_diff

and is assumed to be t-distributed with dof degrees of freedom.

Parameters:¶

value1float or ndarray

Value, for example mean, of the first sample.

value2float or ndarray

Value, for example mean, of the second sample.

std_difffloat or ndarray

Standard error of the difference value1 - value2

dofint or float

Degrees of freedom

alternativestr

The alternative hypothesis, H1, has to be one of the following

• ‘two-sided’ : H1: value1 - value2 - diff not equal to 0.

• ‘larger’ : H1: value1 - value2 - diff > 0

• ‘smaller’ : H1: value1 - value2 - diff < 0

difffloat

value of difference value1 - value2 under the null hypothesis

Returns:¶

tstatfloat or ndarray

Test statistic.

pvaluefloat or ndarray

P-value of the hypothesis test assuming that the test statistic is t-distributed with df degrees of freedom.