Dear Agnieszka,
A.A.Reid wrote:
> Hi Will,
> Thank you so much for the advice on correcting for multiple comparisons!
>
> I am sorry to bother you, but would be very grateful for advice on Design
> Matrix and contrasts which can be specified on the 2nd level analysis in
> SPM5.
>
> I have fMRI data from an event-related experiment with 3 conditions
> (words, pseudowords & baseline). The baseline was not explicitly modelled
> in my design matrix at the 1st level. I have two groups of participants:
> an experimental group and control group. I have run the 1st level analyses
> obtaining con images.
>
> I am interested in comparing the two groups using 2nd level analysis on:
> 1) the activation for Words. I have specified and (estimated) my design
> matrix using the two sample t-tests and obtained the following DM (please
> see the attachment in Microsoft Word). I haven’t managed to find an
> example of a two-sample t-test in the SPM5 manual and just would like to
> double check whether this one looks o.k? I guess what worries me here is
> that I cannot see, listed on the right, all the ‘con’ images which I have
> entered. I can see 9 ‘con’ images for my Experimental group (with ‘D’ at
> the end of the file names), but I am expecting to see 18 images.
> Similarly, I cannot see all the ‘con’ images for the Control group (I can
> see 8, but I am expecting to see 15). Strangely, however, under ‘Design
> Description’ (at the bottom
> of the page), all my 33 con images are mentioned. Is my DM o.k.?
>
Its fine. SPM does'nt list all the data files as the font size would
have to be too small.
> Moving on to the issue of contrasts, I am interested in the following
> questions:
> 1) I would like to know for ‘Words’ which areas were significantly more
> activated in the Experimental group (group 1) than in the Control group
> (group 2). I presume that with my attached DM (or if this DM it is wrong,
> with a DM where the 1st column is labelled Group 1 and the 2nd is labelled
> Group 2), the t contrast 1 -1 would give an answer here. Is this correct?
>
Yes, that should work fine.
> 2) I would also want to know for ‘Words’ the areas that were significantly
> more activated in the Control group (group 2) than in the Experimental
> Group (group 1), then I presume a t contrast of -1 1 would be appropriate.
> Is this correct?
>
yep.
> 3) Finally, I would like to include 3 covariates in my analyses on the 2nd
> level. I have noticed that the SPM5 interface lets me do it with a
> two-sample t-test, which is quite unusual. Usually, in statistical
> packages, one can only use covariates with Analysis of Variance, but not
> with a t-test. Is specifying the covariates in two-sample t-test in SPM5,
> correct?
>
Yes, that's fine. A two-sample t-test is just an ANOVA with two-levels.
very best wishes,
Will.
> Thank you very much for your help.
> Agnieszka Reid
> Aston University, UK
--
William D. Penny
Wellcome Trust Centre for Neuroimaging
University College London
12 Queen Square
London WC1N 3BG
Tel: 020 7833 7475
FAX: 020 7813 1420
Email: [log in to unmask]
URL: http://www.fil.ion.ucl.ac.uk/~wpenny/
|