Dear SPMers,
>> 2.) I ran paired t-test on my previous data set (16 subject, 2
>> conditions, one scan for each condition). The program ran fine, but
>> running the "result" module did not reveal any signal. SPM96 on the same
>> dataset showed very good signal. Does any of you have the same
>> experience? What possibly have I done wrong?
>
> We had a similar problem when trying to run the Linux binaries from
> spm99b - the images displayed well but when it came to the stats part
> there were no results (we had seen results on other platforms). Perhaps
> this is bug?
It seems that I have generally the same Problem. However, I just try to
get started with spm99 analysing a dataset from a single subject (120
scans, 3 conditions, simple motor task in blockdesign).
When I analyse this data using spm96 or other programs I can see very
robust and highly significant activation (corr. P<0.05).
However, when using spm99b (on a Sun) I only see rather weak activation at
an uncorrected threshold of 0.01. This also looks very scattered and is
distributed in a cloud of small clusters.
I can not imagine what could have been wrong. I tried to enter anything
as in spm96 and the designmatrix looked fine, except that the lowpass
filter part was not displayed (what is probably a feature of spm99?).
I also investigated the realigned, normalised and smoothed data
(simple correlational analysis, timecourses,..) and they were really
ok. Therefore, I think it must be some fault in the statistics.
Is there some hidden source of error? I cannot test if this is platform
dependent because I have no access to different computer systems.
Perhaps there is someone out there who could give me a hint?
Thanks,
Christine Preibisch
Klinikum der Universitaet Frankfurt
ZRAD - Institut fuer Neuroradiologie
Schleusenweg 2-16
60528 Frankfurt
Germany
Phone: ++49 69 6301 4651
Fax: ++49 69 6301 5989
email: [log in to unmask]
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|