>
>
>
> Finally... how did it work!? Did you obtain good results? Better/
> worse than a typical cluster inference?
>
Hi, Tom:
Thanks for your help. Obviously, you didn't abandon your friends at
USA though you physically moved to Europe:-).
I have one group of 16 cocaine abusers. The analysis is to assess the
correlation between FA and one treatment outcome.
If I use the regular methods, I did not get any significant cluster
using -c 1.8. If I use -c 1.5, I got big clusters in the bilateral
corona radiata, and other frontal white matter. These clusters are so
big and occupy more than 50% of the corona radiata. This is good and
makes sense. I like such result. But I'm afraid that I can not
report it. Because t=1.5 equals p = 0.1 for two tails test. So, I
decided to try TFCE. TFCE reveal two small, but significant
clusters. I believe each cluster is less than 20 voxels. Though I'm
disappointed for such small clusters, I'm still glad that I got
convincing positive results and I can report them now. One drawback
for use TFCE is that the processing time a little bit too long. I'm
using a brand new Mac with dural processors, with 4 G RAM and 2.4 GHz.
TFCE takes it 3 days for one stats.
> -Tom
>
> ____________________________________________
> Thomas Nichols, PhD
> Director, Modelling & Genetics
> GlaxoSmithKline Clinical Imaging Centre
>
> Senior Research Fellow
> Oxford University FMRIB Centre
|