hi Jef
> I am attempting conduct a paired sample t-test on one group of 10 subjects
> that were scaned 2 separate times; 1 prior to training and 1 post training.
>
> At this point I have attempted to enter each subject's pre and post scans as
> a pair into a paired sample t-test model. I did this at the second level
> analysis, selecting paired sample t-test. I then set up for 10 new pairs of
> images in which I entered each subject pre and post scans as a new pair.
>
> My question arise as I look at the design matrix this model created. I am
> attempting to create 2 contrast: pre - post and post -pre.
>
> Questions: 1) The conditions are not unique, so is this a problem, did I do
> something wrong in setting up the model.
No, that's fine, and normal when you have columns for subject effects
in your model.
> 2) More importantly, since there are 10 pairs of images (for 10 subjects)
> there are 10 blocks in my design matrix. (See attachment) From what it looks
> like here on the design matric in the attachment, this is not pre - post,
> but instead, subject's 2 pair of images being subtracted from subject's 1
> pair of images. Could someone please help me to sort this out. How do I need
> to set up the model to investigate the contrast of pre images - post images
> and the post - pre.
The way that SPM does the design matrix the columns for the subject
effects (which you don't care about) are first, and the last 2 columns
are your conditions. So instead of:
[1 -1 zeros(1,10)]
which is what you have now, and is, as you say, just testing the first
two subjects, you want:
[zeros(1,10) 1 -1]
Hope this helps,
Jonathan
|