Hi,

I see in this forum that the NPM has now been replaced with NiiStat. As far as I understand, NPM supported the use of the Brunner-Munzel test for creating VLSMs, but in the NiiStat introduction it states that "[r]esults are identical to a Student's pooled-variance t-test (if one of the dimensions is binomial)". Does that mean that NiiStat does not support the Brunner-Munzel test?

TIA

Lars Eirik

(PS. I tried asking about this in the NiiStat help forum, but didn't get an answer, so I thought I'd try this forum, as it seems more active.)

True, NiiStat no longer includes the Brunner-Munzel test. The results are therefore GLM (e.g. where the mean is the best measure of central difference). In general, the Brunner-Munzel test is a bit more sensitive if the assumptions of the GLM are violated (I like to think the BM test generally detects differences in median). However, both provide robust permutation based thresholding for family wise error, and NiiStat includes the ability to model nuisance regressors. I do think that NPM is robust, I am simply spread too thin to support a lot of these older projects. All the source code is open source, so any one can extend or maintain these legacy tools.