 |
 |
David Sharp wrote: > Dear SPM'ers, > > We are struggling at present with a random effects analysis. > > We have 2 groups: gp 1 n=12, gp 2 n=8 with an ABC study design. We are > interested in comparing (B+C)-A across the 2 gps to determine if there is > significant difference between the activations. > > Our hypothesis is that GP 2 will activate the homologous region in the right > hemisphere compared with activation on the left for normal controls (gp1). > > We have a 2 main questions: > > (1) For the 1st stage analysis which model should be used. Can we use 'Multi > subj conditions x subj interaction and cov' or does it have to be explicitly a > multi group model. If we have to use a multigroup model can we use 'multi-group > conditions and covariates' or only the 'Full Monty' and what determines which of > these is valid. > In the two-stage RFX analysis, the subjects are placed into groups at the second level - when you do the two sample t-test you tell SPM which con*.img's are from the first group and which from the second. Therefore your first option - 'Multi subj conditions x subj interaction and cov' - should suffice. > > (2) In the 2nd stage analysis using a 2-sample t-test is it valid to lower the > threshold for significance below p 0.05 corr for regions where a specific a > priori hypothesis is held. > The standard approach here would be to use a Region of Interest (RoI) analysis. This can be implemented using the S.V.C. button in SPM - you define a sphere or BOX defining your a priori RoI around your activation - and SPM will tell you the p-values corrected for multiple comparisons over just this region (rather than the whole brain). Alternatively you could use MRI_CRO to define regions that are'nt spheres or boxes. See the SVC notes in Mathew Brett's tutorial for more details: http://www.mrc-cbu.cam.ac.uk/Imaging/ Hope this helps, Will. > > Thanks > > Catrin Blank and David Sharp > > Dr David J Sharp, > MRC Cyclotron Unit, > Hammersmith Hospital, > Du Cane Road, > London, > W12 0NN > Tel: +44 (0)20 8383 3733 > Fax: +44 (0)20 8383 2029 > [log in to unmask] -- William D. Penny Wellcome Department of Cognitive Neurology 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/