FYI-
In my experience, the outlier detection procedures can quickly become problematic as you add inputs and EVs. For example, earlier this summer I ran a large analysis (310 inputs, 8 "true" EVs + 31 dummy EVs for subject means, 7 copes into F-tests) with just FLAME-1, and it finished in a little over 4 hours. After looking at some of the gfeat report's line plots, I wanted to run the analysis with the outlier detection module on. I killed the analysis after 10 days. From what I could tell from the tmpcopeXXXX outputs, it was just shy of 2/3 finished. IMHO, outlier detection is best suited for smaller tests like single group means. Those tend to finish in less than 2 days.
-Jim Porter
[log in to unmask]
On Sat, 20 Aug 2011 14:28:10 +0100, Maria Witek <[log in to unmask]> wrote:
>Hi!
>
>I am doing higher-level FEAT analysis with outlier detection. I started running it about 24 hr ago and it's still not done. The FSL documentation says that outlier-detection "increases computation time considerably", but how much longer should I expect it to take? (I have 26 1st-level FEAT directories).
>
>Using the top command, I can see that the computer is running flameo and taking up 100% CPU.
>
>Hope someone can help! Thanks
>
>All best,
>Maria
|