This is the Python implementation of the McNemar's mid-p statistical significance test for comparing the predictions from 2 models.
While statsmodels
provides the functionality for the McNemar's test, it only implements the exact version of the test and not the mid-p version. The latter is provided by MATLAB in the testcholdout()
function, and this repository aims to implement the same in Python.