Hello Donald,
Thank you for your very informative answer (as usual)!
Yes, a (Bonferroni) correction dependent on the number of post-hoc contrasts is a must. My problem with small volume correction is that it appears to focus on a cluster of specified size and shape centered on a maximum-effect voxel of choice. I would instead like to correct for the entire mask where I find a significant F statistic. Wouldn't it make sense to recompute the smoothness, and FWE corrections, by focusing only on the mask where a significant F was observed?
Cheers,
Bruno
Sent from iPad
On 2012-02-03, at 7:04 PM, "MCLAREN, Donald" <[log in to unmask]> wrote:
> Bruno,
>
> Here are some thoughts:
> (1) You could use small-volume correction -- this will correct for regions that were significant. This method uses the pooled error term and spatial smoothness of the data to your advantage.
>
> (2) A statistician might say you should correct for the number of post-hoc tests, but people have argued that since you have many more brain voxels that comparisons, correcting for tests isn't necessary. However, I still think you need to correct for the number of tests.
>
> (3) If we combine these two thoughts. For regions that are significant, we want assess what are the pair-wise differences. Since we are only testing those regions, it makes sense to use small-volume corrections (+correcting for the number of tests). If you use a cluster extent and cluster p-value, you could use any number of mutliple-comparisons for correcting for the number of tests. I never got a good answer to whether the voxel p-values should be corrected for the number of tests or if the clusters p-values should be corrected for the number of tests. Doing voxel correction changes the results.
>
>
> Best Regards, Donald McLaren
> =================
> D.G. McLaren, Ph.D.
> Postdoctoral Research Fellow, GRECC, Bedford VA
> Research Fellow, Department of Neurology, Massachusetts General Hospital and
> Harvard Medical School
> Office: (773) 406-2464
> =====================
> This e-mail contains CONFIDENTIAL INFORMATION which may contain PROTECTED
> HEALTHCARE INFORMATION and may also be LEGALLY PRIVILEGED and which is
> intended only for the use of the individual or entity named above. If the
> reader of the e-mail is not the intended recipient or the employee or agent
> responsible for delivering it to the intended recipient, you are hereby
> notified that you are in possession of confidential and privileged
> information. Any unauthorized use, disclosure, copying or the taking of any
> action in reliance on the contents of this information is strictly
> prohibited and may be unlawful. If you have received this e-mail
> unintentionally, please immediately notify the sender via telephone at (773)
> 406-2464 or email.
>
>
>
> On Fri, Feb 3, 2012 at 12:29 PM, Bruno L. Giordano <[log in to unmask]> wrote:
> Hello,
>
> I am analyzing a factorial design. After FWE correction, I obtained a significant main effect of a 3-levels factor (F test). I would like to run some post-hoc contrasts within the significant-F region (a few clusters), to test for pairwise differences between the various levels of the design factor.
>
> My question: is it appropriate to carry out the post-hoc contrast by using the same FWE correction (computed considering the entire brain) when my interest is in a rather restricted area? Don't I run the risk of being too conservative and/or imprecise?
>
> Can someone please suggest the standard approach in this case? Should one re-estimate the proper FWE correction by re-running the factorial model within the regions of interest?
>
> Cheers,
>
> Bruno
>
