 |
 |
Dear Philipp, It is possible (but difficult) to set up a paired t-test GLM for FIRST, as you must be careful to have the EVs demeaned, and arrange it so that the contrast you would normally specify (as FIRST doesn't allow contrasts to be specified at the moment - although we plan to change this in the future) is a simple contrast with only one non-zero value (e.g. [0 1 0 0 0 0]). You then just use the output from FIRST that is associated with the EV which would have had the non-zero value in the contrast. This result would then represent the main paired t-test result. The vectors for this (if it is a single group) are easy to interpret, but not so when there is more than one group. Even if you get this right it isn't an ideal solution however, as you don't take into account the paired nature of the images in the fitting process (something we are working on). Hope this helps to clarify my previous response. All the best, Mark On 5 May 2012, at 16:43, Philipp Thomann wrote: > Dear Mark, > > wouldn't it be ok to compute the F-statistics and to identify the direction of changes as indicated by the arrows in FSL view? > > Best regards > Philipp > > > > Am 05.05.2012 um 02:51 schrieb Mark Jenkinson: > >> Dear Luke, >> >> I'm afraid we are still working on a longitudinal version of vertex analysis. >> You can't implement a paired t-test with the current FIRST tools. >> However, you could save out the vertex coordinates using the --saveVertices >> option and then use other statistical packages to implement a multivariate >> paired t-test. >> >> Hope that helps. >> All the best, >> Mark >> >> On 5 May 2012, at 05:49, Luke wrote: >> >>> Dear FSL-Experts! >>> >>> I wonder if FIRST can be used for longitudinal studies ( in my case 2 groups (Placebo / Intervention) examined at two points (T1/T2)). The analysis of the subcortical volumes via fslstats already revealed the expected result. Now I would like add a vertex analysis to infer shape changes due to the on the effect of the intervention. Is there any issue of theoretical or practical kind that might argue against performing a vertex analysis in this situation? Any recommendations for the first_utils options in this situation? >>> >>> Best wishes and many thanks in advance, >>> >>> Luke >>> >