 |
 |
Hi Eugenio, Please see below: On 31 January 2018 at 11:06, Eugenio Abela <[log in to unmask]> wrote: > Hi Anderson > > two basic follow-up questions I have on this: > > - I want to control for disease duration as a EV, which is obviously only > present in patient groups. Can I still include this in one design (i.e. > controls would have all zeros in this EV), or should I use separate designs > (as you show in the PALM examples page)? > You can include the disease duration as an extra EV, setting 0 for the non-patient groups. > - What is the correct reference for the -corrcon option, i.e. the method > you use to identify the dependence structure of the contrasts? > The reference for correction over contrasts is: Winkler AM, Webster MA, Brooks JC, Tracey I, Smith SM, Nichols TE. Non-parametric combination and related permutation tests for neuroimaging. Hum Brain Mapp. 2016 Apr 5;37(4):1486–511. http://dx.doi.org/10.1002/hbm.23115 All the best, Anderson > > Thank you very much > > Eugenio > > On 29 Jan 2018, at 13:29, Anderson M. Winkler <[log in to unmask]> > wrote: > > Hi Eugenio, > > Thanks for commenting. The NPC options aren't necessary here. You'd > assemble a design that has the EVs coding for the different groups > (patients and its subgroups, controls), prepare the 6 contrasts, then run > as usual just with the option "-corrcon" included. > > It isn't the same as Bonferroni but if, incidentally, the contrasts happen > to be all independent from each other, then the result will happen to be > equivalent to Bonferroni. > > Hope this helps. > > All the best, > > Anderson > > > On 29 January 2018 at 04:33, Eugenio Abela <[log in to unmask]> > wrote: > >> Dear Anderson, >> >> Thank you for making PALM freely available - I'm new to it, and the first >> few test runs have worked like a charm! >> >> The problem I'm facing now is how to adequately correct over multiple >> contrasts. I'm running an analysis in which I will need 6 pairwise >> comparisons (Patient vs controls, patients with sydrome 1 vs syndrome 2, >> patients with mild disease vs patients with severe disease - and vice >> versa). All use the same modality, EEG images from SPM12. >> >> As I understand, this can be achieved in PALM using the option -corrcon, >> which does FWER over contrasts. Is this a Bonferroni correction? If so, >> power might be an issue, and I would want to try contrast-level correction >> with NPC. >> >> How does the latter work in practice? Would I need to use -npccon >> -corrcon as command option? And in terms of interpretation, would it mean >> that partial contrasts are thresholded using the joint FWE threshold, or >> this incorrect? >> >> Thanks for any help! >> >> Eugenio >> > > >