Dear Georgios,
you can try to make your regressors more normally distributed by applying log or exp function depending on the distribution. Sometimes also polynomial functions will help. Simply try some of these functions and check the transformed distribution with the hist function in Matlab. If your data are closer to a normal distribution, you are on the safer side even if the deviation from normality might be not that important as Tom pointed out.
Best,
Christian
On Thu, 2 Aug 2018 11:51:52 +0100, Georgios Argyropoulos <[log in to unmask]> wrote:
>Dear Experts,
>
>Apologies if the question is naive. I was wondering if it is possible to
>run whole-brain VBM regression (on modulated GM) using between-subjects
>regressors of interest (e.g. behavioural scores) that are not normally
>distributed (within the context of analyzing structure-behaviour
>relationships).
>
>If not, would you be able to recommend any alternatives?
>
>many thanks for your time
>
>Georgios
>--
>Georgios P. D. Argyropoulos, PhD, CPsychol, AFHEA
>http://argyropoulos.blogspot.co.uk/
>Postdoctoral Researcher in Memory Neuroscience
>Medical Sciences Division
>Nuffield Department of Clinical Neurosciences (NDCN)
>Level 6, West Wing, John Radcliffe Hospital
>Oxford
>OX3 9DU
>UK
>
|