Hi Henk,
Yes, there is. Just edit the *.con file, add the new contrasts that you
want and delete those you've already run, and edit the field
"/NumPoints" to reflect the new number of contrasts. Move the previous
results to a separate directory to avoid overwriting them, and run
randomise again. The new outputs will be named *tstat1, *tstat2, etc,
according to the new contrast file. You may want to rename them to
continue the original sequence and avoid confusion.
If you also have new F-tests, use a similar strategy, configuring the
*.fts file according to the new *.con file. If the new tests are only
F-tests, make the changes, and run randomise with the option "--fonly".
All the best,
Anderson
Am 18.02.14 09:56, schrieb H van Steenbergen:
> Dear FSL-ers,
>
> I was wondering whether there is a way to quickly run some additional contrasts using randomise without rerunning the complete model again. In my case I am running a VBM analyses with more than 400 subjects and about 20 contrasts, so knowing how to "add" a few contrasts without running all the others again might save me a couple of days computational time.
>
> If I could simply calculate the non-parametric reverse contrast of existing contrasts (e.g. HC > PATIENTS based on an existing PATIENTS > HC contrast), that would also be fine already. I was used to standard parametric statistics before using randomise. With standard statistics one can simply use positive and negative z-scores, but this is obviously not going to work when using permutation-based testing given the directional testing and the resulting p-values.
>
> Thanks a lot for your help in advance.
>
> Best, Henk
|