Dear all,
I'm a newbee in SPM and I'm in charge of solving a recurrent problem in the last versions of SPM5 (the first one isn't concerned as far as my tests are valids :)).
You can find, attached with this message, 2 screenshots of the same simple t-test "1 vs all" analysis on PET images ("essai-apres.jpg" after having updated my version of SPM5, "essai-avant.jpg" before, ie with the first version of SPM5).
One solution I found on this forum is to specify the Variance factor to "Equal", instead of the default value "Unequal". But I guess, thanks to the SPM5 manual, that's not a valid way to treat the problem (the parameters of my analysis are sampled quite differently).
I also checked a lot of matlab versions without any success (my current one is Matlab2008).
I don't want to ignore the possibility of a bad use by "my" SPM users, so could anyone tell me which factor(s) could tell upon the behavior of this statistical computation?
Is it then possible to bring back the first version of SPM5 (June 2010) from the web site or another ftp site?
Thanks a lot for your answers in advance,
Have a nice day,
Fabien
|