On Thu, 1 Nov 2007 13:10:30 +0800, Zhang Zhiqiang
<[log in to unmask]> wrote:
>Hi SPMers,
>
> I'd like to know whether it's OK to do 2nd level analysis using
spmT*.img instead con*.img. what's the difference between them?
con* are the contrast images. They're the images of the actual effects.
Meaning: they're linear combinations of the estimated regression
parameters, the beta* images.
spmT* are the t-score maps corresponding to the con* images. They're
essentially maps of significance.
If you read the SPM documentation (see the link on the main SPM website),
you'll see that the way SPM does "random effect analysis" (the group-level
analysis that yields inferences that are valid for the entire population
from which subjects are drawn) involves a two-stage method. First models
are created and estimated for each subject, then images coming from those
subject-level (first-level) models are brought to the group level (second
level), where they're used as data in group level models.
The reason con* and not spmT* images are used as data in the second level
models is simply that that's the correct way of doing it. Meaning that if
you look at how to do random effects analysis in a textbook---which has
the appropriate definitions, and presumably does things in one stage, not
two stages---and then look to see how a _correct_ two-stage method works,
you'll see that correct two-stage methods use con* not spmT*.
>
> Thank you very much!
>
> ----------------------
> Zhiqiang Zhang
> Lab. Neuroimage, Dep. Med. Img
> Clinical School, Nanjing University. China.
>
>
>---------------------------------
> @yahoo.cn 新域名、无限量,快来抢注!
|