I'm running a 4-variable analysis in FIRST (one group, three covariates), and I have significant results with the F-test. I now want to see the direction and size of the effects; in a typical model I would look at the betas/model coefficients, but these are not available with the vertex analysis. I see in the documentation (http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FIRST/UserGuide#randomise) the recommended approach is to do a t-test, or look at the actual values with the FSL timeseries tool (or by extracting the data).
The problem with the t-test is that I only get one result ("....tstat1") even though I have 4 contrasts (see design.con below), and I get 4 F stats. I'm not sure which contrast tstat1 refers to, and furthermore I'm not sure how to translate the t statistic for this 4-variable model to an effect size. For a t statistic from a 2-sample t test multiplying by std/sqrt(N) would work - is it the same with a 4 variable model?
Similarly, the issue with extracting the individual values is that I while I could look for group differences, I cannot account for the 3 other variables in the model without running another analysis on the extract values, which is possible but no ideal.
So my questions are:
- is it possible to obtain the betas?
if no,
- how can I obtain the t statistic maps for my 4 contrasts?
- is there a way to translate the t statistic in this 4-variable model to a non-standardized effect size?
Any guidance appreciated!
-------------------
design.con
/ContrastName1 group mean
/ContrastName2 TIVvol mean
/ContrastName3 age mean
/ContrastName4 sex mean
/NumWaves 4
/NumContrasts 4
/PPheights 1.552420e+00 1.719039e+00 9.529518e+01 1.722726e+00
/RequiredEffect 0.673 0.673 0.560 0.417
/Matrix
1.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00
0.000000e+00 1.000000e+00 0.000000e+00 0.000000e+00
0.000000e+00 0.000000e+00 1.000000e+00 0.000000e+00
0.000000e+00 0.000000e+00 0.000000e+00 1.000000e+00
|