Hello FSL users,
I'm currently trying to wrap my head around entering a demeaned RT at the individual level. I have found a couple posts already in the archive but I'm struggling to apply them to my own experiment/analysis so I was wondering if someone could help clear up a few details?
In my experiment we have a 3 (decision) x 2 (category) self-paced event-related design. Items are presented in blocks of 6 trials with one null event per block. This results in 6 EV's (three column format; onset, duration (individual RT), and weight set to 1) which are used to develop the FEAT design matrix. By using RT as the duration in the model I (hopefully!) have accounted for time on task differences between my conditions, however even while controlling for duration, I'm interested if there are differences in the height of the response.
Having had a look at the archives I found this post - https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind1102&L=fsl&F=&S=&P=31660
This post suggests that I should enter 2 regressers in the design matrix; the same one I have already used (onset, RT, weight) and a second EV whereby the weight is set to the demeaned RT for that experimental task (I think!) for that participant.
Now I'm unsure of how I set up my contrasts at the individual level; at the moment at the individual level I only model the contrasts of the conditions over baseline/rest which I have also modelled explicitly using the three column format. In my old analysis the contrast of EV to, rest, and null events is simply 2 - 1 - 1.
But now I have 2 EV's for each condition I'm unsure how to model it. Is it simply a case of giving both EVs' (original and demeaned) a value of 1 and leaving the rest/null events at -1? Or do I need to orthogonalise the two EV's and use that in my baseline contrast?
Any help would be greatly appreciated.
Thanks,
James D
|