Li, Yu wrote:
> Dear Will:
> Thank you very much for your guide on our last question.
> We have few other questions. In our fmri study, each subjcet(depress patient) at baseline has 2 sessions in 1 run.
> Secession 1 : ABABA (A: Look at Neutral pictures) (B: Look at negative pictures)
> Secession 2 : CBCBC (C: Look at Positive pictures) (B: Look at negative pictures)
> Both are box-car.
> We get very good activation map of A<B (with contrast -1(A) 1(B) ) from Secession 1,
> but for secession 2 C<B (with contrast -1(c) 1(B) ) we only see very few activation in all subjects.
> Could you please give any advice what may be the reason for secession 2.
Dear Julia,
From a 'methods' point of view you have done nothing wrong.
You correctly conclude that Negative-Neutral gives you many activated areas
but Negative-Positive gives you few activated areas.
It's not my area of research but might you be interested in
areas that show any change - ie positive or negative ? In
which case you could do an F-contrast using -1 1.
> Dose it makes any difference that set (-2 1) instead (-1 1)?
Yes. This would be asking a different question: show me all
voxels where activity in condition B is more than twice that in
condition C (instead of just where it is more than in C).
So I don't think you want this.
> I also tried to do cross secessions that compare A and C, and didn't see activation in the map for A<C or A>C.
> The contrasts for A>C that combined 2 secretion ABABACBCBC is 1 0 -1 0. For A<C is -1 0 1 0. Is it correct setting?
Yes. Assuming your conditions are ordered ABCB, -1 0 1 0 is the correct contrast to use.
Again, you may want to use this with the F-contrast option to show two-sided
effects.
> If we want to compare ABC. B as baseline, how to set up contrast?
Again, if conditions are ordered ABCB to look at the difference between within-session
effects you'd use 1 -1 -1 1 (ie. from (1 -1) - (1 -1)). Again, you may wish
to use the F-contrast option if you have no idea which sign the effects will
take.
> Thank you very much for your help and time.
>
Very best wishes,
Will.
> Julia
>
> -----Original Message-----
> From: Will Penny [mailto:[log in to unmask]]
> Sent: Tuesday, September 30, 2003 4:58 AM
> To: Li, Yu
> Cc: [log in to unmask]
> Subject: Re: covariate
>
>
>
>
> Li, Yu wrote:
>
>
>>Dear Will:
>>In fMRI, Spm99. For within a patient group at baseline and patient group before and after treatment
>>
>
>>how can we use the anxiety ratings as a nuisense covariate and depression ratings as a covariate of interest.
>>
>
>>Specifically, how do we set up the contrasts?
>>Thanks
>>
>>Thanks
>>Julia Li
>>
>>
>
>
> Julia,
>
> Let's say you have 4 patients.
> Then you could create a design matrix, X, with 8 rows (corresponding to the 8
> scans you have - before and after for each subject) and 8 columns
> corresponding to the variables (1) before, (2) after, (3-6) subject 1 - 4
> effects, (7) anxiety and (8) depression.
>
> The matrix should look something like this
>
> X^T =
> [1 1 1 1 0 0 0 0;
> 0 0 0 0 1 1 1 1;
> 1 0 0 0 1 0 0 0;
> 0 1 0 0 0 1 0 0;
> 0 0 1 0 0 0 1 0;
> 0 0 0 1 0 0 0 1;
> 3 2 1 4 1 2 4 3;
> 10 20 30 10 1 2 7 4];
>
> where the ^T denotes transposed - so one row of the above matrix should
> correspond to a column in the design matrix image.
>
> You should be able to create this in SPM using the
> PET design option: Multi-subj (conditions & covariates) with
> 4 subjects and 2 conditions -
> enter the condition variables [1 2] for before and after, and
> enter the covariates by hand - or from variables you've loaded into
> the MATLAB workspace.
>
> Then, to look at the main effect of treatment use the
> contrast [-1 1 0 0 0 0 0 0]. To look at the
> main effect of depression use [0 0 0 0 0 0 0 1].
>
>
> Note that the 8 'scans' you have will be *con* images from a
> first-level fMRI analysis. All of the above is a description of
> your 'second-level' analysis.
>
> Alternatively, if you use differential contrasts from the
> first level eg. con = after-before, then you'll only need one
> 'condition' at the second level.
>
> Hope this helps.
>
> BEst wishes, Will.
>
>
>
>
>
--
William D. Penny
Wellcome Department of Imaging Neuroscience
University College London
12 Queen Square
London WC1N 3BG
Tel: 020 7833 7478
FAX: 020 7813 1420
Email: [log in to unmask]
URL: http://www.fil.ion.ucl.ac.uk/~wpenny/
|