Hi,
I'm currently having a hard time understanding how SPM uses the beta values, residuals and noise variance in order to calculate the spmT maps. More specifically, I have several long MRI sessions from the same participant on a short sequence - roughly 15 sessions with 3000 volumes each, per participant. Concatenation in 1 GLM seems hardly possible, it's very time consuming to estimate or even open an SPM.mat (and often makes matlab crash due to memory overload).
Therefore I am summing up the beta images (using imcalc) across the separately estimated GLMs for the same subject (as you would do with concatenation). Now I'm trying to figure out a correct statistical approach to estimate the t values from these summed beta images. Any ideas? I thought about correcting the summed betas by dividing them with the square root of summed product of each session's ResMS x xX.Bcov. Might that be the right direction?
cheers,
|