I have question and I’m sure someone knows the answer. I am revising a paper and addressing a reviewer’s concern. I used CAT12 for longitudinal VBM to compare a placebo and treatment group and the reviewer has a concern about the use of parametric testing (flexible factorial group x time model, post-hoc t-tests) and the assumptions of normality and equal variance. I know that smoothing addresses the Gaussianity of the data, but is there a specific resource/reference that addresses the issue of potential between-group heteroscedasticity in the data?
Hi,
I might be misunderstanding your question but:
I don’t think this is what the reviewer had in mind. Yes, within a participant there is a gaussian smoothing kernel, but that doesn’t mean the data across subjects are normally distributed. I imagine this was the point of the reviewer, since models are run in individual voxels across subjects.
None that I can think of off the top of my head, especially without knowing about your sample sizes (e.g. larger samples are likely going to be more normal, if the two groups are large and about the same size then they may also have similar variances). Is there a reason why you cannot directly test these assumptions or use non-parametric tests?
Best,
Steven