This is just straight-forward GLM, ie, beta = inv(X'*X)*X'*y (y is the input time series after preproc) yhat = X*beta residual = y - yhat residualstd = sum(residual.^2)/DOF  (DOF = rows of X - cols of X) t = C*beta/sqrt(C'*inv(X'*X)*C*residualstd) where C is the contrast matrix

On 3/10/2023 4:50 PM, haoran.x@gmail.com wrote:

External Email - Use Caution

Hi,

I have a naïve question. When I use GLM for a single subject timeseries data, how the variance is evaluated to get the t value? I know the beta values and contrast matrix make the numerator for the t value. How is the denominator made? Is it through the fitting residual ie std(residual), or is it from session by session or run by run variance? Can I find the formula somewhere? I was not able to find the answer in tutorial or the existing mailing list. Thanks a lot!

Best,

Haoran

---

Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer