Hi,
We’re running a higher-level analysis using 244 inputs
and 4 EVs (we’re just doing an ANOVA as per the guidelines on the FSL
website). We used FLAME1 with outlier de-weighting, and the following errors
appear repeatedly in our logs:
fdtr domain error
ndtri domain error
There aren’t any other errors, and the same analysis
works without errors when we do not implement the outlier de-weighting feature.
Curiously, despite the domain errors, we still get results, although I suspect
these results can’t be trusted (?).
Is this simply a memory issue? We’re running our
analysis on and SGE cluster and each node has 4 CPUs and 8 GBs of RAM (and we’re
using 64bit).
Thanks,
David