Dear Margaret,
I'm afriad I haven't got time at the moment to answer all three of your
questions, but I think there might be a problem in your basic random
effects
analysis: This probably arises from a confusion of terminology,
particularly
contrast images (weighted sum of parameter images), which is different to
the resulting statistic image (i.e. SPM{T}). Here are my thoughts:
At 14:08 17/06/99 -0700, Margaret Zhao wrote:
| (3) we can perform a t-test on SPMt*.img(SPMz*.img) to generate
| pseudo random effect analysis, or even we can average these
| SPMt*.img (SPMz*.img) and assume the averaged data with mean=0
| and std=1/sqrt(n) for a quick look. Is there any way similar so
| that we can directly deal with SPMF*.img to see the group effects?
To effect a standard random effects analysis on a balanced design you would
perform a t-test on the *contrast* images (con_????.img), not the
*statistic* images (spmT_????.img).
Performing a test on the statistic images is essentially testing whether
there is significant reliability of detection of an effect across subjects,
rather than testing the effect itself.
Averaging Gaussianised statistic images and assuming mean=0 and
std=1/sqrt(n) is essentially combining evidence of an effect from each
subject in a meta-analysis across *these* subjects, and is not a random
effects analsis, as no account is taken of the variability of effect across
condition.
This was discussed in more detail in July last year, so have a look at the
following thread if you want further discussion: (The last two messages are
the most relevant.)
http://www.mailbase.ac.uk/lists/spm/1998-07/0026.html
http://www.mailbase.ac.uk/lists/spm/1998-07/0027.html
http://www.mailbase.ac.uk/lists/spm/1998-07/0028.html
http://www.mailbase.ac.uk/lists/spm/1998-07/0032.html
http://www.mailbase.ac.uk/lists/spm/1998-07/0034.html
Hope this helps,
-andrew
PS: Jean-Baptiste ot I will get back to you on your real questions soon!
+- Dr Andrew Holmes ------------------ mailto:[log in to unmask] -+
| ___ __ ___ Robertson Centre for Biostatistics |
| ( ,) / _)( ,) Boyd Orr Building, |
| ) \( (_ ) ,\ University Avenue, |
| (_)\_)\__)(___/ Glasgow. G12 8QQ Scotland, UK. |
+----------------------------------------- http://www.rcb.gla.ac.uk/ -+
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|