Skip to content

P-values incorrectly reported for two sided tests. #23

@estebanz01

Description

@estebanz01

Some two sided tests are reporting wrong p-values, hence wrong hypotesis evaluation results. According to https://support.minitab.com/en-us/minitab/18/help-and-how-to/statistics/basic-statistics/supporting-topics/basics/manually-calculate-a-p-value/ we should calculate the p-value with the following rules:

  1. if test is assuming a lower tail, p-value = CDF(t_score). We show this calculation as probability.
  2. if test is assuming a greater tail, p-value = 1 - CDF(t_score). We show this calculation as p_value for one_tail tests.
  3. if test is assuming a two sided test, p-value = 2 * ( 1 - CDF( |t_score| ). We dont calculate the CDF of the absolute value of the t_score. We use the normal t_score. This is yielding weird results reported in the p_value field.

Metadata

Metadata

Assignees

No one assigned

    Projects

    Status

    Done

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions