Dear Lynn,
> I have a few elementary questions about statistical inference:
>
> 1) Many people reference Bailey et al (1991) in setting a height threshold
> of p < .001, uncorrected, to reduce risk of false positives with an omnibus
> test. Should one assume that this is adequate and that, having rejected
> correction, one should use the SPM96 default extent threshold of p = 0.5?
There was a detailed discussion of the use of the results of this abstract
on the list in April last year. A search in the archives for 'Bailey'
should pull it out.
> 3) Under what, if any, circumstances would it be acceptable to use a less
> stringent height threshold, possibly with the extent setting of 0 as
> suggested by Dr. Friston's previous note of Oct.5? For example, what about
> when there are anatomical constraints on the basis of an a-priori
> hypothesis, or when an independent measure such as a manually placed region
> of interest calculation gives statistically significant results?
You could try my utility, vol_corr, which gives adjusted p values for
Z values within a specified volume of interest:
ftp://ftp.physiol.ox.ac.uk/pub/matthew.brett/Corr_p
This implements the equations by Keith Worsley that Christian mentioned
in his email. There is a readme file there that discusses the issues a
little bit.
Cheers,
Matthew Brett
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|