 |
 |
At the risk of exposing my total ignorance, I would like to ask for some advice about interpretation of results of SPM99 analysis of PET and SPECT data. Can you recommend any papers which describe (preferably in language intelligible to a non-statistician) how to decide (i) what height and extent thresholds to set (ii) if the height threshold should be for p uncorrected or corrected (iii) then, from the list of reported regions, whether to report as 'significant' only regions which have p<0.05 (or whatever) at the voxel level or only those which have p < whatever at the cluster level (iv) whether the p values used in stage (iii) should be corrected or uncorrected I had understood that the purpose of the corrected p was to correct for the difference between the much higher probability of a significant statistic at any voxel in the brain, and the much lower probability of significance at any one specific voxel, so that existence of a prior hypothesis about the location of an effect would be justification for using uncorrected p values to examine for an effect at that location. In the literature, I see that people typically use thresholds of p<0.05 or p<0.01, but beyond that there seems to be a lot of diversity for comparable types of studies as to whether people used p corrected or uncorrected (even without prior hypotheses), and this, as well as other details, including the height and extent thresholds, which can impact on the regions listed, are sometimes not specified at all. Any comments on this apparently confusing situation? Any help would be appreciated. Nanette -- Nanette Chana Freedman, Ph.D. Hadassah University Hospital Department of Medical Biophysics and Nuclear Medicine Kiryat Hadassah P.O.B. 12000 91120 Jerusalem, ISRAEL Tel: 972 2 6776 720 (work) Fax: 972 2 6421 203 (work) Tel: 972 2 5636 174 (home) Email: [log in to unmask]