Hi Joe,
> Dear SPM community, we used the FDR method on one of our data sets,
> setting the FDR parameter to 0.0001. A colleague of mine is
> suggesting that we report the thresholding this way:
>
> p < 0.0001, FDR corrected
This works, though I think
> FDR < 0.0001
is clearer. But I think
> q < 0.0001
is probably best avoided, since "q" hasn't become universal (different
authors have laid claims to different definitions of "q values").
Also, if we use "q" for FDR, should we use "r" for FWE? And what if a
new metric is created.
Best bet is to state clearly the type of FDR used in the methods
section (currently, saying you use SPM's FDR is sufficient, b/c there
are not yet multiple flavors... they're coming though!) and then use
something like "P_{FDR} < 0.05" or "FDR < 0.05" (where _{} indicates
subscript).
-Tom
-- Thomas Nichols -------------------- Department of Biostatistics
http://www.sph.umich.edu/~nichols University of Michigan
[log in to unmask] 1420 Washington Heights
-------------------------------------- Ann Arbor, MI 48109-2029
|