Message

Hi All, I wanted to clarify a point. Publications of cortical thickness differences are usually produced with maps called mean thickness difference maps.

I dug through the mail archive but did not find an answer

(Q1)

These statistical parametric maps are I assume produced by computing a mri_glmfit with doss, and a contrast (1 -1 0) i.e. each class is

fit to a linear model but both have the same slope and the offsets are different and the inference is on the difference in the intercepts. The resulting maps are the maps commonly referred to as mean thickness difference maps ?

(Q2) I seem to remember a conversation about how this is done. First each group is fit to an optimal line, i.e. each group gets a line with the intercept

and slope to minimize the data fit error, then the average of the slopes is computed and this average slope is used for each of two new lines, one for each group, then an intercept is computed for each group such that with the nonoptimal slope the data fit error is minimized, am I correct that this is in fact

the procedure.

(Q3) Does this in fact give an intercept that is the mean of the thickness values for each group , the thought experiments I have cooked say yes but

I have not made a proof ?

Thanks

Greg