Rik:
Many thanks for answering so quickly.
On Thu, 02 Apr 2009 11:38:53 +0200, Rik Henson
<[log in to unmask]> wrote:
>
> Apologies, but I don't know precisely why you have such strange results.
> Are you sure you entered the scans correctly (ie didn't select repeat
> scans by mistake)? BTW, you also have a very weak design, with only 16
> scans and hence 6dfs (rank(x)=10). Did you try to esimate nonsphericity?
> If not, I think there must be another mistake somewhere, because I have
> never seen such errors.
We have found such strange results in 3 larger data sets. This design was
only a small part of one analysis for demonstating the problem.
On nonsphericity:
in flexible factorial following specifications were made
1st factor: name: subject; independence: yes; variance: equal
2nd factor: name: Gruppe; independence: yes; variance: unequal
3rd factor: name: Bedingung; independence: no; variance: unequal
main effect: 1
interaction: 2 3
>
> BTW, I may not be the best person to help, because I don't use SPM5's
> 2nd-level GUI (if you want batch scripts for general within/between
> ANOVAs, try this
> http://www.mrc-cbu.cam.ac.uk:8080/people/rik.henson/personal/batch_spm_anova.m),
> and haven't yet matured to SPM8, so cannot answer your question about
> what has changed.
I've run the script with nsph_flag = 1 and found negative F values again
for the contrast
1 1 -1 -1 .5 .5 .5 .5 -.5 -.5 -.5 -.5
The absolute F values were different from those obtained by the flexible
factorial GUI or by SPM8. There are also some differences in the designs
(as you may see in the attached files).
BTW: I've just read that SPM8 is released. Should we proceed simply with
SPM8?
Best regards
Bertram
>
> Rik
>
> Bertram Walter wrote:
>> Dear SPMers, dear Rik,
>>
>> I have used flexible factorial for a 2 (between subjects) by 2 (within
>> subjects) factorial design. Each group consisted of 4 subjects. This
>> resulted in the following design:
>>
>> First 8 columns are subject's constants, last 4 columns regressors for
>> the repeated measures in the two groups.
>>
>> 1 0 0 0 0 0 0 0 1 0 0 0
>> 1 0 0 0 0 0 0 0 0 1 0 0
>> 0 1 0 0 0 0 0 0 1 0 0 0
>> 0 1 0 0 0 0 0 0 0 1 0 0
>> 0 0 1 0 0 0 0 0 1 0 0 0
>> 0 0 1 0 0 0 0 0 0 1 0 0
>> 0 0 0 1 0 0 0 0 1 0 0 0
>> 0 0 0 1 0 0 0 0 0 1 0 0
>> 0 0 0 0 1 0 0 0 0 0 1 0
>> 0 0 0 0 1 0 0 0 0 0 0 1
>> 0 0 0 0 0 1 0 0 0 0 1 0
>> 0 0 0 0 0 1 0 0 0 0 0 1
>> 0 0 0 0 0 0 1 0 0 0 1 0
>> 0 0 0 0 0 0 1 0 0 0 0 1
>> 0 0 0 0 0 0 0 1 0 0 1 0
>> 0 0 0 0 0 0 0 1 0 0 0 1
>>
>> The contrast for the comparison of groups is
>> 0.5 0.5 0.5 0.5 -0.5 -0.5 -0.5 -0.5 1.0
>> 1.0 -1.0 -1.0
>>
>> Using an F test for this contrast gave an spmF image with negative F
>> values.
>>
>> As the test compares only two groups I calculated also a T test for
>> this contrast. The resulting spmT image contained only zeros.
>>
>> What was wrong? Using spm8b for the same design and tests gave
>> reasonable results.
>>
>> BTW: The beta maps for the first 8 regressors (subjects) were slightly
>> different in spm5 and spm8b. The remaining 4 beta maps seem to be
>> indentical. What has been changed in the model estimation process?
>>
>> Cheers and
>> thanks in advance
>>
>> Bertram
>>
>
--
Dr. Bertram Walter
Bender Institute of Neuroimaging
University of Giessen
Otto-Behaghel-Str. 10H
35394 Giessen
Germany
Phone +49 (641) 99-26307
or +49 (641) 99-26331 (Secretary)
Fax +49 (641) 99-26309
|