 |
 |
And secondly, whether spm has functions to perform lowpass filtering? (Only highpass is present as an option in the gui) On 22/05/2009 19:19, "MCLAREN, Donald" <[log in to unmask]> wrote: > Alex, > > (a) I would make sure that the seed time-course regressor is orthogonal to the > nuisance covariates. To do this, simply use the regress function in matlab > with your seed as the DV and the nusiance variables as the IV, the residuals > are what aren't explained by the nusiance regressors. Use the residuals > instead of the timecourse. > > (b) your contrasts are fine; however, the contrasts represent the slope of the > relationship between the seed region and other voxels. If you want the > correlation coefficient, then you need to convert the T-statistic to an > R-squared value AND for group analysis convert that to a normal distribution > using the Fisher Z-transform. > > (c) if you haven't subtracted the global mean from your data, you won't see > the anti-correlations and need higher group thresholds to limit the regions > that you find. > > On Fri, May 22, 2009 at 5:05 AM, Alex Fornito <[log in to unmask]> wrote: >> Hi all, >> I'm relatively new to spm and would like some clarification re: how multiple >> regression is implemented. >> I am trying to set up a 1st level design, which is a seed-based correlation >> analysis of resting-state data. As such, there is no design matrix (in the >> sense of a task regressor), and so instead I have created a matrix of 10 >> timecourse regressors, which I have loaded in using the multiple regression >> option. >> The first column in the matrix is the timecourse extracted from a seed >> region, and is the regressor of interest. The other 9 columns in the matrix >> are nuisance covariates (e.g., timecourses from white matter, csf, movement >> parameters, etc). >> After estimation, I'm presuming the contrast [1 0 0 0 0 0 0 0 0 0] will >> return all voxels positively correlated with the seed timecourse, whereas >> [-1 0 0 0 0 0 0 0 0 0] will return those negative correlated with the seed. >> >> My questions are: >> (a) does this sound like the correct way to set up the analysis? >> >> (b) would the results returned by the contrasts defined above be >> automatically corrected for shared variance with the other 9 nuisance >> regressors, or do they require some kind of contrast weighting as well to >> ensure they are covaried for? >> >> Thanks for your help, >> Alex > > -- Alex Fornito CJ Martin Post-Doctoral Fellow Brain Mapping Unit Department of Psychiatry University of Cambridge Downing Site Downing St, Cambridge UK CB2 3EB Email: [log in to unmask] Phone: +44 (0) 1223 764670 Fax: +44 (0) 1223 336581 Australian Details: Melbourne Neuropsychiatry Centre National Neuroscience Facility Levels 1 & 2, Alan Gilbert Building 161 Barry St Carlton South 3053 Victoria, Australia Email: [log in to unmask] Phone: +61 3 8344 1861 Fax: +61 3 9348 0469