Hi Tim,
Here's another statistician wondering about tbss inference.
I have two specific questions:
1. What is the difference in using -c 2 or -c 3 in randomise?
If this is related to the cluster size, as you write, how does
2 differ from 3? What is the unit?
If I overlay the results of these two runs, I get somewhat more
significant areas by using 3 than 2. Some of them are overlapping,
others not.
I've gone through some of the papers by Nichols but I haven't
found any satisfactory answer.
2. In your technical paper you either 'correlate' covariates with
FA or 'regress them out'. What does this mean literally? Do you
compute voxelwise correlation coefficients (and if so, how do you
get them? How do you know whether they are positive or negative, large
or small? Are they ordinary correlations for linear dependence or some
nonparametric versions?).
Does 'regressing out' mean the usual 'adjusting for'? How do you
get the
coefficients of the covariates if GLM type estimation is used?
Is there any paper you would recommend to clarify these otherwise
rather standard statistical ideas in this context?
Does 'correlating' technically mean that the covariates are in the
original design matrix whereas 'regressing out' means that they are
technically put to another design matrix with the prefix -x?
Thanks in advance for any clarification.
Best wishes,
Mervi
|