Testing normality with respect to version and software used decides usage of non-parametric test. The first tests, MATLAB (v1) and v1, as a whole, failed to be considered nornal meaning testing for significance across all versions and softwares should be done by non-parametric tests, only!

For comparing counts...

Means:

Validation:

V1: 88.8989%
V2: 88.7363%
V3: 89.2668%

Test:

V1: 88.8795%
V2: 89.4053%
V3: 89.5274%

MATLAB:

Valid:

90.0378

Test:


90.5089

SAS Viya:

Valid:

84.9533

Test:


84.6275