Hi Anath
> I performed the first-level analysis two ways on a block design task, 4
> epochs of 24 seconds each interspersed between 5 epochs of 30 seconds each,
> with a TR of 3 seconds, for a total of 82 measurements/volumes.
> First, I performed the analysis without the motion correction parameters
> giving me a total of exactly two of those beta hdr/img pairs.
> Then I included the motion parameters from the Realign step. That gave me a
> total of 8 beta_xxxx.hdr/img” files. There are 6 motion parameters. Along
> with the condition I defined, that gives a total of 7.
> Is the last parameter the error? One beta for the condition, one beta for
> every motion correction parameter, and one for the error. That would make sense.
The last parameter corresponds to the overall mean (the last column in
your design matrix, which is just a constant).
> Secondly, I set the units for design to “seconds” instead of the default
> “scans”. I defined the condition in seconds, but the motion parameters are
> in scans. There are 82 rows. When I add the rp_*.txt file to the “multiple
> regressors” field of the “Subject/Session” added to the “Data & Design”
> portion, does it automatically assume that it is in scans? Will this mess my
> analysis up? If not, what can I do to convert it to seconds? Because it
> would be easier for future when I convert everything to seconds.
The way that "regressors" are handled in SPM during the first level is
to have one value per scan, so these can be added as-is (regressors
are not convolved with a basis function, like the hrf, so are treated
somewhat differently than "conditions" in SPM). You can use the
"multiple regressors" option to enter all the movement parameters in
one go.
> Finally, I am very confused regarding the 2nd level analysis. Assuming each
> beta is for one condition, if I had the betas of several control subjects
> and patients for the same condition, I want to get the group activation maps
> for both the patients and the controls. And I also want to find the
> differences in the activation between the two groups. What do I use for
> that? One-sample t-test, 2-sample t-test, paired t-test, multiple regression
> or flexible factorial? When would I use each design in the 2nd level
> analysis? I want to do random effects analysis. I’ll appreciate any help.
The simplest approach would be:
-To get the mean activation for condition A for controls -> 1 sample t
test on beta*/con* images for controls
-To get the mean activation for condition A for patients -> 1 sample t
test on beta*/con* images for patients
-To compare patients and controls, 2 sample t test (and assume
variances are not equal) on those same images.
Hope this helps. Good luck!
Jonathan
|