Hi @rsmith ,
I am running into a similar problem with my design and contrast matrices as I am working with three groups in my experiment (AD, aMCI, and controls) and I was wondering whether fixelcfestats still cant test more than one hypothesis at a time? My second study (allele based) only involves one hypothesis but anytime I include columns for my covariates (normalised age, sex, ICV) it crashes. Any guidance would be greatly appreciated!
Hi @H.A.N.K,
There’s nothing precluding you from running multiple hypotheses. The contrast matrix can have one row for each hypothesis as a t-test, and you can then separately nominate multiple F-tests.
The more interesting question is whether you should be testing more than one hypothesis. Is it truly the case that you have six independent hypotheses:
, or are you rather interested to know whether these groups are different, ie. reject the null hypothesis that all groups are equal? Because that can be done with an omnibus F-test akin to an ANOVA. And if you don’t mind git branch hopping and compiling from source, it’s possible to then perform explicit post hoc testing to interrogate which of the group differences drove the statistical significance of the F-test in a way that preserves the false positive control guarantees of the original test. This is what we used in this work, for which the implementation is here.
As for command crashing, you would need to provide your exact command invocation, the resulting error message, and an adequate description of the input data.
Cheers
Rob