Hi
> I'm just wondering what the outputs mean (I know about corrected and uncorrected results). My direct question is what are exactly dr_stage3_ic00xx_tfce_corrp_tstat1 , dr_stage3_ic00xx_tfce_corrp_tstat2 , dr_stage3_ic00xx_tfce_corrp_tstat3 and dr_stage3_ic00xx_tfce_corrp_tstat4?
> Which statistical map are they and what is their difference?
>
They are (1-p)-value maps ('corrp') after TFCE correction that relate to a specific original IC being tested for significance of connectivity difference under a GLM model with a set of 4 different contrasts. tstat1 is the (1-p) value map for GLM contrast 1. If one or more of these statements do not make much sense to you then you should go back to the online FSL course to learn about GLMs and contrasts
> As the instruction says, I show those images in FSLVIEW and adjust threshold between 0.95 and 1 . In this case, I see nothing in dr_stage3_ic00xx_tfce_corrp_tstat1 and 2 but I see functional connectivities (similar to meaningful networks) in dr_stage3_ic00xx_tfce_corrp_tstat3 and 4?
That's possible, yes. I suspect your design is a simple 2 group unpaired t-test design? In that case your 'interesting' contrasts are 1 and 2 (as they look at between group differences, assuming you sued a canonical glm_gui setup) and contrats 3 and 4 check for within group mean. If I'm right then you have nothing and points (i) to (v) below become unnecessary
> (I have 60 components and in all of them I see the same thing as described above)
> What should I with these results?
>
You should do multiple things:
(i) check that your processing is OK (seeing the same effecta cross 60 components suggests that it;s not)
(iI) understand what they mean (procedurally)
(iii) correct for the number of comparisons (60 in your case), i.e. threshold at 1-0.05/60 to 1
(iv) see what survives and interpret findings biologically
(v) if you're confident about having done a good job in (i)-(iv) then write up, submit and face the scrutiny of the community
hth
Christian
> Thanks very much,
> Karen
|