> In a posting to the SPM e-mail group, you wrote:
>
> The two p-value formulas are quite different. The Gaussian random
> field one is, in D dimensions:
>
> P-value proportional to RESELS * z^(D-1) * exp(-z^2/2) (approx)
>
> The `Bonferroni' one would be
>
> P-value proportional to RESELS * z^(-1) * exp(-z^2/2) (approx)
>
> As you can see, the formulas are different functions of the
> threshold, z, so they can never agree, even if you try to
> adjust the constants of proportionality.
>
> I'm confused by the use of RESELS in the Bonferron equation. Do you mean
> VOXELS instead?
Jerry,
Nope, I meant RESELS; the second one *shouldn't* make any sense.
It represents a case where you do a Bonferroni correction based on the
RESEL count (which is wrong). The above text shows that doing a RESEL-based
Bonferroni p-values doesn't equate to the random field based p-values.
Does this help?
-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
|