Subject: | | Re: Multi-masking for Multiple Comparison Correction |
From: | | Roberto Viviani <[log in to unmask]> |
Reply-To: | | [log in to unmask][log in to unmask]] > Sent: Thursday, October 02, 2008 9:25 AM > To: Fromm, Stephen (NIH/NIMH) [C] > Cc: [log in to unmask] > Subject: Re: [SPM] Multi-masking for Multiple Comparison Correction > > I'm actually with Roberto on this one. In all these cases, we're using > inferential statistics. The validity of the *inference* that we make > based upon the statistics is the important thing in this case. If we > clearly state that the inference is limited by the apriori > assumption/condition that we made, then we shouldn't have a problem. > > Take the trivial case of a functionally defined mask based on a > contrast A that is used to mask the same contrast, as mentioned by > Stephen. Obviously the inference "in this region of the brain, A was > significant" would be non-valid. But if instead we made the inference > "in regions of the brain where A was significant we were able to show > that A was significant" we would be absolutely fine, statistically and > inferentially speaking, though a reviewer would question our sanity in > finding it worthwhile to report. > > A more realistic example: I perform a study in which I show people > angry and happy faces and black dots. I define a contrast face-dot and > find regions of the brain showing this effect. I use those regions as > a functionally-defined ROI and test the angry-happy contrast. What > inferences can I make? "In regions showing greater activation for > angry and happy faces than for dots, we found angry faces to produce > greater activation than happy faces". This would be fine I think. But > it would be wrong to drop the first part of that statement. > > -Tom > > On Thu, Oct 2, 2008 at 1:56 PM, Fromm, Stephen (NIH/NIMH) [C] > <[log in to unmask]> wrote: >> Roberto, >> >> Sorry if this isn't addressing your points; the original poster's >> question was a little unclear to me, because I wasn't 100% sure what > she >> meant by "multi-masking." >> >> All I mean is that if you use a mask defined by the same functional > data >> that you're applying the mask to, the significance will possibly be >> inflated. >> >> As for definitions, an example which is somewhat conceptually related > is >> stepwise regression. The paper at >> http://publish.uwo.ca/~harshman/ssc2006a.pdf >> states, "When model modifications are selected using post-hoc >> information (e.g., in stepwise regression) standard estimates of >> p-values become biased." Ultimately, I think my use of "bias" here is >> correct, based on definitions given at Wikipedia. >> >> So, what I'm saying here is that the use of a EGä‰t |
Date: | | Fri, 3 Oct 2008 18:21:27 +0200 |
Content-Type: | | text/plain |
Parts/Attachments: |
|
|
|
|
Independence is a concept applicable to random variables. In the
frequentist approach, violations of the null hypothesis are fixed (not
random). In the example you make, the null hypothesis is violated by
the angry faces condition, but the errors are still random. The test
statistics in question remain independent assuming normality of the
errors. If the normality assumption is violated, the test statistics
are only uncorrelated.
As a first reaction, I'd say that the problem in the example you
present is not where you think it is. The ANOVA assumption is that
your fixed effects sum to 0, so you will not be able with your ANOVA
model and your contrast to determine that it's only angry faces that
drive the association, only that the angry - happy difference is
positively associated with the signal.
But say you'd included neutral faces. Then you could use pairwise
comparisons, and I see no problem with the case you mention where you
find that only angry faces differ from neutral. But notice the
conditionality on the first contrast: your design includes only two
types of faces, this larger design three, so that the set of voxels
examined in the second contrast may be different.
- Roberto
<snip>
> The condition of orthogonality seems to me not entirely clear. Is it
> the contrast vectors that need to be orthorgonal, or the contrast
> estimates themselves? e.g. In my example of a face and dot
> presentation task, my functional ROI might be determined using:
>
> angry faces: 0.5
> happy faces: 0.5
> dots: -1
>
> whereas my final contrast of interest is:
>
> angry faces: 1
> happy faces: -1
> dots: 0
>
> These two contrast vectors are orthogonal. But if the first contrast
> is driven exclusively by angry faces producing high activation and
> dots and faces producing little activation, is the ROI defined on that
> basis truly orthogonal to the contrast being tested? Because in the
> extreme case, it would seem to collapse to the situation in which the
> mask used to restrict the search space wholly corresponds to the
> contrast being tested. Would a more appropriate approach using my
> example be to define a functional ROI based on a conjunction of
> angry-dots and happy-dots contrasts?
>
> -Tom
>
|
|
|
|