comparison between PLS and SPM
as22kk
Posted on 08/14/08 09:29:47
Number of posts: 60
Hello PlS experts,
I am in the final stage of using multivariate method that I have faced the strange result(to our opinion) by comparison of PLS and SPM5 analysis. I have got 2 sinificant LVs from PLS analysis and I was planning to set the same contrast from LV1 for spm analysis(random effect analysis) hoping that the main effect of availabilitu signal shows a similar spatial pattern to that identified in task PLS.For that ,I put the special focuse on those regions which reach the highest bootstrap ratio during LV1 and search (by setting the same contrast value as LV1)in SPM if I can find some of them with reseanable distance(10 mm) and acceptable Z value,But SPM mostly could not find any of those regions(Thresholds for the statistical maps were set at P <0.001 uncorrected to facilitate comparison with the PLS results.).By the way,I have to state that all the regions which have been detected by SPM were also detected by PLS.
Furthermore,I have been using all the preprocessed data which had been produced using SPM5,so the preprocessed data for both type of analysis was exactly the same.
Any idea will be appreciate.
Thanks,
/Alireza
Untitled Post
rmcintosh
Posted on 08/14/08 09:37:14
Number of posts: 394
Check out the paper:
McIntosh AR, Chau WK, Protzner AB (2004) Spatiotemporal analysis of event-related fMRI data using partial least squares. Neuroimage 23:764-775.
it does a a comparison with SPM and indicates why there are differences in the results. In your case, it may be partly to do with sensitivity and partly to do with the use of a modeled HRF.