On Sat, Feb 22, 2014 at 5:42 AM, Alessandro Calamuneri <[log in to unmask]> wrote:

Hi Donald and n.,
sorry for writing in your conversation, I would like to understand a point about Donald suggestion on comparing each subject separately to the healthy group.

How you can perform that within spm?

You would perform 4 two-sample t-tests for each condition. In the two-sample t-test, you'd enter all the control subjects as one group and the patient as the second group. Set the variance to be equal between the two groups as their is only 1 variance measure. Then estimate the model and form the contrast of 1 -1 or -1 1 depending on your hypothesis. This will show you where the patient is different from the controls.

Would it work to procede as follow, namely considering beta images for a given contrast, taking the mean of the contrast for control group and comparing such mean with the sambe beta for each patient?

No, you don't want the mean image of the controls, because you need to know the variance of the controls.

Maybe this is not correct as you do not model within subject variance..

Otherwise would it work using a second level analysis inserting only all betas for one contrast in one group and the same contrast from each subject separately in the other, running then a 2 sample t-test?

Yes. I believe this is what I described above.

Thanks
Alessandro