fMRI plotting residuals

Hi everyone,

I am interested in trying to understand GLM model fitting better in fMRI. I really like the approach of visualizing residuals after running a model. I have implemented my own version of this tutorial in NiLearn, but I have pretty much no frame of reference for what is a reasonable deviation of non-normality. In the tutorial, of course, the residuals are all distributed around zero. Mine are… not, to say the least. Does anyone know of any discussions that have taken place where people talk more about their model fits and residuals in fMRI?


you can check residuals normality with scipy:
However, it may be more important to monitor the correlation in the time domain.

Thanks so much for the insights :slight_smile:
I use scipy.stats.normaltest right now, but I can give the Shapiro test a shot. Regarding your comment about correlation in the time domain, do you mean temporal autocorrelation?

Yes, temporal autocorrelation. Bertrand

1 Like