Dunnett vs. Bonferroni post hoc on ANCOVA

anagrammarian

I am doing some ANCOVA analysis on thalamic vol differences between 4 groups (CN,EMCI, LMCI, AD) of controls vs. stages of Alzheimer's using age and ICV (intracranial volume) as covariates. When doing posthoc, I am getting radically different results using Bonferroni and Dunnett. My Bonf. results match what I get using Dunnett on XLSTAT. Any thoughts? It seems like the means used for Dunnett seem to be uncorrected (corr to Descriptives below) whereas Bonferroni seems to use corrected (corr. to Marginal means below). I would think the latter should be used for both. And from clinical and other evidence the Bonferroni results make sense.

Thanks,

manoj saranathan

JohnnyB

Hi @anagrammarian ,

Thanks for pointing this out. I think it will be good to at least provide a footnote to these tables that states that the Dunnett (and the other non-standard post hoc tests) are based on the uncorrected means. In the future we will add the Dunnett based on the corrected means (it does seem to be an obscure test though, and I cannot seem to immediately find an R-package that does this).

Kind regards

Johnny