> Keith Worsley has mentioned the problems of using a 'RESEL Bonferroni'
> correction, but I'm not sure if he has ever written anything about it.
> In [1] we reported the problem in math mode in section 3.2.3, and with
> simulations showed how it was very anticonservative (rejected too
> often) for all smoothness and df considered (see Fig 6).
>
> [1] Nichols & Hayasaka
> Controlling the familywise error rate in functional neuroimaging:
> a comparative review
> Statistical Methods in Medical Research, 2003; 12: 419-446
>
Obviously I need re-reading papers I last perused too long ago ... I
certainly haven't heard Worsley mentioning this so it must have come from
your paper.
What I do remember, however, is that you mentioned there that the bootstrap
hadn't been investigated for the purposes of the maximum distribution in
random fields. I wonder if you went ahead with it. We tried several
bootstrapping schemes as part of our ADC map investigation, but I am not
sure that if we wrote them up we would be reporting anything new.
Roberto Viviani
Dept. of Psychiatry
University of Ulm
Germany
|