Dear Sukhi
>
> Thankyou for your quick reply.
> I'm using a two condition experimental design - repeated ABABAB etc. I
> tried to place the y as a regressor and got an error message specify user
> specified regressor. So the problem may be a bug as you suggest.
Did you already download all the bug correction ? I experienced this bug on
another computer than mine that may have not been updated, when using the
'specify a model' option.
>
> The secondary issue is the group - academic at this point as i cannot get
> the program to accept the regressor! But how can i specify the "good" part
> of y for each subject?
The best is to cut your y vector into pieces. Each pieces should have the
length of each of your sequence. The best way to manage this is to reshape y in
matlab with the first dimension naming the sequence number, and the second for
the scan number within the sequence (but this is only if all your sequences
have the same number of scan), so you will only have to specify y(1,:), y(2,:)
.... The other way to do that will be to specify the number by hand : i.e.
y(1:100), y(101:200) ... if your sequence length are of 100 scans. You will
have to specify this for all your sequences.
>
> I have no objection to you posting your response to the mailing list. I
> felt more comfortable e-mailing one person than littering many e-mail
> intrays crossthe world! I am, as you can see, very much a novice
This was also my case not many time ago ;-)
May I further comment on your design, assuming that you will try to perform a
PPI. The canonical approach for that, is to use a factorial design (F1+F2+,
F1-F2+, F1+F2-, F1-F2-). You can than use an area corresponding to the
activation of one factor say F1, and test for the interaction of this area with
F2. The point is that mathematically, your regressor should be orthogonal (in
another word, all the variance explained by the area of interest and F2 should
be regressed from the interaction). However, you raise an interesting question
that I would address to the SPM experts :
Let say that you do this : You have only one factor (F1) and you use the area
A1 corresponding to that factor as 'F2'. Assuming that you regressed all the
variance of F1 and A1 from your interaction term (F1xA1), will the result be
valid ?
In fact this is a continuation of the question that Richard Perry and I raised
some mails ago :
Say that A is correlated with the factor F1
Say that the PPI of A*F2 (a second factor) gives B.
The question is wouldn't we have to check that F2*F1 does not also fit B ?
Good luck for you analysis
Hope that could be helpful
Jack
_________________________________________________________________
| Jack Foucher Universite Louis Pasteur |
| Institut de Physique Biologique UPRES-A 7004 du CNRS |
| 4 rue Kirschleger Tel: 33 (0)3 88 77 89 90 |
| 67085 STRASBOURG Fax: 33 (0)3 88 37 14 97 |
| France |
| Faster E-mail: [log in to unmask] |
| Other [log in to unmask] |
|_______________________________________________________________ |
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|