Dear Matthew and all,
In perhaps loose relation to the purpose of your new code,
I wonder if a "continuity correction" for SPM results has been
determined yet. That is, SPM can give overly conservative results
if the level of smoothing is not greater than some minimal level
(even more conservative than Bonferroni correction in some cases;
personal observations and personal communication from Keith Worsley).
This seems to me to be the main disadvantage of general use of SPM
results (however theoretically beautiful they might be). I wonder
if the need for a continuity correction becomes even more of an issue when
dealing with non-convex, small search volumes such as those relevant
to the unified SPM result. Also, I recall the FIL group mentioning
in some abstracts about a possible solution to this problem by
oversampling the data and then smoothing, thus allowing more valid SPM
results for small levels of smoothness relative to the original
spatial resolution. Has this issue been explored/discussed in
greater detail anywhere?
Thanks,
Eric
>
> Dear SPMers,
>
> There have been several questions recently on how to derive corrected p values
> and thresholds when looking at smaller volumes than the whole brain - e.g.
> when you have an apriori hypothesis. It has I think been mentioned that
> the formulae for such corrected p values were published by Keith Worsley
> et al in 1996, Human Brain Mapping 4:58.
>
> This is just to let you know that I have now on ftp put some code which
> implements these formulae to give corrected p values and thresholds:
>
> ftp://ftp.physiol.ox.ac.uk/pub/matthew.brett/Corr_p
>
> There is a README file there. The archive is corr_p.tar.gz.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|