Hi AFNI experts,
I am using different basis functions to see which one better models my data. In particular, I'm using TENT and CSPLIN, varying timepoints covered/number of functions/baseline polort.
1) What is the advantage of having individual basis function as interger and/or TR-locked?
2) When calculating the area under the curve using TENT or CSPLIN, which one should you choose 1/2Bo+B1+B2+B3..+1/2Blast or 0*Bo+B1+B2+B3..+1/2Blast? Or is it better to have a subset of timepoints that are likely to cover the peak of the response?
3) I have 4 runs each lasting about 660s; should I really use -polort 5? (I have been using 2 but have been recommended to use 5 every time I run 3dDeconvolve)
4) In general, what are the key things that you should look into when comparing different models? I have compared the stats, including full F-stat and R2, individual cofficients and their associated t-stats and baseline cofficients. I also compare whether activation seems to rise and fall within the window of time I designated. Is there anything else I should look for the comparison of different basis functions?
5) There obviously are a lot of intra/inter-individual differences in this type of exploration, and so how many subjects should I examine in order to decide upon a "relatively better" model to continue with for the next level of group analysis?
6) One of your regression manauls indicates that the CSPLIN do not produce statistically different IRF estimates than the TENT (only its shapes are more smooth); then what is the advantage of modelling with the CSPLIN functions?
Thank you so much for your help!