The same rules should apply. However, its not clear that dividing by
the number of tests you are doing is the right way to analyze the data
as you are already correcting for multiple comparisons across voxels.
The question arises, do you correct at the voxel or cluster level for
the number of tests or do you say you have N-times the number of
voxels and do the correction that way?
Best Regards, Donald McLaren
=================
D.G. McLaren, Ph.D.
Research Fellow, Department of Neurology, Massachusetts General Hospital and
Harvard Medical School
Postdoctoral Research Fellow, GRECC, Bedford VA
Website: http://www.martinos.org/~mclaren
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 Tue, Nov 27, 2012 at 2:52 AM, Yann Quidé <[log in to unmask]> wrote:
> Thanks Donald for your reply.
>
> Just a little methodological question to follow up. I thought one has to
> correct for the number of post-hoc tests if using 2-sample t-tests in the
> ANOVA model. Is it the same rule with separate 2-sample t-tests (with less
> degrees of freedom)?
>
> Thanks.
>
> Yann
>
>
> On Tue, Nov 27, 2012 at 4:36 PM, MCLAREN, Donald <[log in to unmask]>
> wrote:
>>
>> On Wed, Nov 21, 2012 at 8:16 PM, Yann Quide <[log in to unmask]> wrote:
>> > Dear Experts,
>> >
>> > I was wondering what can be a correct and accepted way to report an
>> > one-way between group ANOVA results.
>> >
>> > Let's say we have 3 different groups. To assess if there is any
>> > difference between them, we performed an one-way ANOVA (F-contrast [1 -1 0;
>> > 0 1 -1]).
>> >
>> > Here is my first question: is there a rule for thresholding this kind of
>> > analysis? Is a threshold of p=0.05 FWE too conservative?
>>
>> >>> No. As long as you correct for multiple comparisons, the method really
>> >>> doesn't matter. Some people use voxel-wise FDR/FWE, some people use
>> >>> cluster-wise FDR/FWE, and some people use 3dClustSim in AFNI to establish
>> >>> extent thresholds.
>>
>> >
>> > Our results show a significant cluster (even with a p=0.05 FWE
>> > threshold). In order to know what is the direction of this effect, we used
>> > separated (3) between-groups two-sample t-tests. In order to be consistent,
>> > we used the same threshold.
>> >
>> > Is that a good way to perform such an analysis?
>>
>> >>>> Yes.
>>
>> >
>> > Or would it be better, and less conservative, to use a different
>> > threshold, lest's say for exemple(and always trying to correct for multiple
>> > comparisons):
>> > - F-test: p=0.0001 uncorrected and reporting only clusters with a
>> > significant peak with FWE correction
>> > - post-hoc two-sample t-tests: p=0.0001 uncorrected at the voxel level
>> > but FWE corrected at the cluster level
>> >
>> > Does it make sense? Is there any methodological issue?
>>
>> >>> I'd use the same thresholding method for both the F-test and post-hoc
>> >>> t-tests. Some researchers would also suggest that you correct for the number
>> >>> of post-hoc tests.
>>
>> >
>> > Thanks for your comments/suggestions.
>> >
>> > Cheers,
>> >
>> > Yann
>
>
|