New subject: Lingering questions for Longitudinal spatiotemporal LME analysis

22 Apr 2024


      External Email - Use Caution
Hello,
I've been following the FreeSurfer longitudinal LME analysis tutorial (
https://secure-web.cisco.com/1UXjrD54pEj2u6JH6Y-Z-dHoSgY6qcbXqPg43uVU2GGyBvm...), which
has been excellent; however I have a few lingering questions pertaining to
my specific analysis. For reference, I am interested in assessing rates of
longitudinal atrophy in each of my two groups (0=controls, 1=clinical).
Regarding what I've currently done:
1). My Qdec file contains 4 columns: fsid, fsid_base, time(years), group
2). I followed the steps to create the appropriate M matrix, and since my
preliminary analysis is simply to create a *group x time* linear model, my
design matrix is X = [ones(length(M),1) M M(:,1).*M(:,2)]; and my contrast
is CM = [0 0 0 1]
3). I've followed the pipeline for the lme_mass_fit_EMinit,
lme_mass_RgGrow, and lme_mass_fit_Rgw functions.
4). I'm skipping the model comparison steps, as I prefer my model with the
two random effects (intercept, time).
5). Get stats (F_lhstats = lme_mass_F(lhstats, CM), applied multiple
comparison correction ([detvtx,sided_pval,pth] =
lme_mass_FDR2(F_lhstats.pval,F_lhstats.sgn,lhcortex,0.05,0); and saved the
sided p-values.
My questions are:
a). Since I'm interested in both hemispheres and the tutorial only assesses
the left, do I need to do step #3 with the right hemisphere data?
b). Would I need to do the last step in the tutorial to get a single
threshold for both hemispheres and then perform the analysis this way?
P = [ F_lhstats.pval(lhcortex) F_rhstats.pval(rhcortex) ];
G = [ F_lhstats.sgn(lhcortex) F_rhstats.sgn(rhcortex) ];
[detvtx,sided_pval,pth] = lme_mass_FDR2(P,G,[],0.05,0);
pcor = -log10(pth)
Thank you for the clarification.
Dan