Hi Carlos,

It should be fairly straightforward to include EVs for your covariates at the first level, as demonstrated here in the FEAT guide.  Differences betwen correct and incorrect responses could be accounted for either by modelling them using separate EVs, or by having one EV model the average response, and another EV with +1 for correct responses, -1 or incorrect responses, modelling the difference between correct and incorrect responses.  

Eugene

2009/5/21 Carlos Faraco <[log in to unmask]>

Dear FSL community,

I am interested in correlating RTs and # of correct responses with fMRI
activation. For the task in question we performed two runs, so I had run a
3rd level analysis to get the group activation map.

From this map I picked the local maxima of interest, dilated them (3mm
sphere), and multiplied these masks by the 2nd-level z-stat image (the
normalized zstat image of the combined runs). The values in these masks were
averaged and I ended up with an average intensity value for each ROI for
each person (the reason for dilating the local maxima from the group map was
because the location of activation in each individual varies, of course). I
input all these values along with the avg # of correct responses and avg RTs
into SPSS and ran a correlation.

My question is, well, is this correct? I read one post that said I should be
using the cope images. Furthermore, should I instead just be specifying
these RT and correct response values into FSL in the 1st level or 2nd level
analysis to run the correlations? If so, I am not really sure how I would go
about setting this up. I have looked through several posts, but they
unfortunately haven't been much help to me.

Thanks,

Carlos Faraco

--

Eugene Duff

FMRIB Centre,
University of Oxford
John Radcliffe Hospital, Headington OX3 9DU  Oxford  UK

Ph: +44 (0) 1865 222 739  Fax: +44 (0) 1865 222 717

--