Hi Xi,
To be clear: You have a clinical group and a control group. You use MVPA
to predict group membership from brain activity patterns and find a
couple of brain regions where local patterns of brain activity are
predictive of group membership. You then perform a two-sample t test to
compare the two groups.
If I understood this correctly, I'd say that it is double dipping
because the "contrast vectors" between the univariate test and the
MVPA-based approach are the same: clinical group v control group. The
only difference is that you started with a multivariate analysis first
(which is sensitive to univariate effects) and then do a univariate one
afterwards.
Best,
Michael
On 29/11/17 14:18, Xi Yu wrote:
> Dear Roberto,
>
> Thanks very much for your quick reply! The first analysis were
> multivariate pattern analyses (MVPA), not one-sample t test. The second
> analyses conducted in regions derived from the MVPA were two-sample
> ttest. Do you think it is still double-dipping?
>
> Thanks,
>
> Xi
>
> <https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail&utm_term=icon>
> Virus-free. www.avast.com
> <https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail&utm_term=link>
>
>
> <#DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>
>
> On Wed, Nov 29, 2017 at 8:12 AM, <[log in to unmask]
> <mailto:[log in to unmask]>> wrote:
>
> Dear Xi,
>
> concerning the ROI analysis, it depends on how you defined the ROIS.
> If you used activation clusters in the one-sample t test to conduct
> regressions on individual differences, it is very unlikely that you
> are doing double dipping. The issue is discussed in
>
> https://www.ncbi.nlm.nih.gov/pubmed/19900563
> <https://www.ncbi.nlm.nih.gov/pubmed/19900563>
>
> Best,
> Roberto Viviani
>
>
>
>
> Quoting Xi Yu <[log in to unmask] <mailto:[log in to unmask]>>:
>
> Dear experts,
>
>
>
> I have a question regarding statistical analyses that I wish to
> get your
> thoughts on. I have two groups, one clinical and one control. I
> first
> conducted multivariate pattern analyses, and found a few regions
> showing
> distinctive activation patterns between the two groups. I then ran
> traditional ROI analyses in each of these regions, and found higher
> activation for the clinical group in one region, higher
> activation for the
> control group in one other region, and no differences in the
> remaining
> regions. Some co-authors are concerned that these analyses are
> double
> dipping, since the second one was conducted in regions derived
> from the
> first one. However, I don?t think so, as you can see from the
> results that
> the first analyses do not associate with any specific pattern in
> the second
> one.
>
>
>
> Could someone please share his/her opinions on this issue?
>
>
>
> Thanks very much for your help!
>
>
>
> Bests,
>
>
>
> Xi Yu
>
> <https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail&utm_term=icon
> <https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail&utm_term=icon>>
> Virus-free.
> www.avast.com <http://www.avast.com>
> <https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail&utm_term=link
> <https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail&utm_term=link>>
> <#DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>
>
>
>
>
>
|