 |
 |
Hi Julian, The within-subject variability is usually much smaller than the between-subject variability such that, although there can be differences in results (with more power where these lower-level are considered), usually we don't see massive differences (as in "robust activations" you mention). FEAT doesn't see the between-subjects dependencies in the case of twins/relatives, and if there are many such pairs in the sample, false positives will be inflated (this isn't a problem with FEAT, it's just how the GLM works, and would happen even in a permutation test that didn't consider these dependencies). All the best, Anderson On Fri, 26 Oct 2018 at 09:16, Julian Macoveanu <[log in to unmask]> wrote: > Dear Tom, > > Thanks for the reply. Actually I already did it in PALM with correct > exchangeability block file, but I get nothing significant, not even > across all subjects. My understanding is that one benefit of FLAME > estimation is that it accounts for within-subject variability and palm > does not since only the cope files from the 1st level are used. Maybe > that could be the reason I got robust activations in FLAME but not > PALM. > > Julian > On Fri, Oct 26, 2018 at 2:11 PM Thomas Nichols > <[log in to unmask]> wrote: > > > > Dear Julian, > > > >> Just wanted to confirm that I got the GLM right. I got 3 groups of > >> subjects: low-risk, high-risk & affected. The participants are twins, > >> most but not all have both twins present. The twin pairs can be in the > >> same group i.e. the low-risk group, but can also be split between the > >> high-risk and affected groups. > >> > >> I have 3 EVs for the three groups and in order to account for > >> within-pair variance correlation I have an extra EV for each twin pair > >> with two 1s for the respective twins pair and 0 otherwise. Would this > >> be an accurate GLM setup? > > > > > > Yes, this approach is 'safe' in that the residuals will be independent. > It's not optimal, though, as you are effective have to toss out any > variance that is common to each twin pair. > > > > If you did this in PALM, you could model this as a simple 1-way ANOVA, > and then tell PALM about the twin structure. But I need to defer to > Anderson to tell you how to hand-craft the exchangeability block file. > Anderson? > > > > -Tom > > > > > >> > >> Cheers, > >> Julian > >> > >> ######################################################################## > >> > >> To unsubscribe from the FSL list, click the following link: > >> https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1 > > > > > > > > -- > > __________________________________________________________ > > Thomas Nichols, PhD > > Professor of Neuroimaging Statistics > > Nuffield Department of Population Health | University of Oxford > > Big Data Institute | Li Ka Shing Centre for Health Information and > Discovery > > Old Road Campus | Headington | Oxford | OX3 7LF | United Kingdom > > T: +44 1865 743590 | E: [log in to unmask] > > W: http://nisox.org | http://www.bdi.ox.ac.uk > > > > ________________________________ > > > > To unsubscribe from the FSL list, click the following link: > > https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1 > > ######################################################################## > > To unsubscribe from the FSL list, click the following link: > https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1 > ######################################################################## To unsubscribe from the FSL list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=FSL&A=1