I'm afraid what you're doing doesn't make sense- copes are (contrasts between) regression coefficients not correlation coefficients.
Also, if you are using one of the non-ols group-level options the varcope is used to feed into the higher level calculations, so you need to leave the cope as it is.
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director, Oxford University FMRIB Centre
FMRIB, JR Hospital, Headington,
Oxford. OX3 9 DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask]
On 29 Sep 2011, at 22:30, Aaron Kucyi <[log in to unmask]> wrote:
> Hi there,
> I ran a first-level resting state fMRI seed-based correlation analysis for a group of subjects using FEAT, and I obtained individual-level correlation maps. If I feed my COPE files into a higher-level group analysis using FLAME, I obtain nice zstat images that look like canonical resting state networks. However, I also tried taking my first-level COPEs and performing Fisher's r-to-z transform on these files before feeding them into a higher-level FEAT (with the intent of ensuring that the correlations are normally distributed). If I do this and select "inputs are 3D cope images" and Mixed Effects Simple OLS, my resulting COPE images are not in correlation units and the zstat images are blank.
> So I have the following questions:
> a) Am I running the higher-level FEAT on Fisher's transformed cope files correctly?
> b) If so (a), is there a way to convert the higher-level stats output files (cope, tstat, zstat) into correlation units?
> c) If so (b), would I expect to see zstat images that are similar to the ones I had obtained with FLAME (without Fisher's transform)?
> Thank you
> Kind regards,