What about, for example, the correlations I've seen in a cohort of subjects.
In 158 subjects aged 10-19 (both controls and patients), the correlation between L & R thalamus is 0.91, and the correlations between L & R of caudate, putamen, pallidum, hippocampus, and amygdala were all 0.75 or higher. I would think that a Bonferroni correction would be incredibly conservative and, in my opinion, just plain wrong because true significant diff's would be missed. Is there any principled way of dealing with multiple tests that are correlated? Thanks, Chris ________________________________________ From: Bruce Fischl [fis...@nmr.mgh.harvard.edu] Sent: Saturday, July 21, 2012 12:01 PM To: Watson, Christopher Cc: Douglas N Greve; freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] Questions about correction over 2 hemispheres and MCC Hi Chris, bonferroni will be overly conservative in that case, but we rarely really know the true covariance structure of the data, so we would rather err on the side of being conservative. cheers Bruce On Fri, 20 Jul 2012, Watson, Christopher wrote: > Hi Doug et al, > > The 2nd question is something I've wondered about. Doesn't a Bonferroni > correction assume that the measures are independent? > If so, I think in the case of subcortical structures, it is incorrect to use > this method, as e.g. the putamen and pallidal volumes are not independent of > one another. > If not, and it is acceptable to use when there are dependencies between > measures, how do you calculate the FWE? It wouldn't be equal to alpha/n, > unless I am misunderstanding something. > > Thanks, > Chris > ________________________________________ > From: freesurfer-boun...@nmr.mgh.harvard.edu > [freesurfer-boun...@nmr.mgh.harvard.edu] on behalf of Douglas N Greve > [gr...@nmr.mgh.harvard.edu] > Sent: Friday, July 20, 2012 2:38 PM > To: freesurfer@nmr.mgh.harvard.edu > Subject: Re: [Freesurfer] Questions about correction over 2 hemispheres and > MCC > > Hi Reem, it looks like you've done everything correctly (sorry for the > null result). As for your second, question, I don't think there is a way > other than Bonferroni. You can try FDR, but it makes the interpretation > a little messy. > doug > > On 07/19/2012 07:59 PM, Reem Jan wrote: >> >> Hi Doug >> >> I hope it?s okay that I double check I?m doing the correction over 2 >> hemispheres correctly and ask a question regarding correction for >> multiple comparisons. I will briefly describe what I have done: >> >> I have 2 groups of subjects ? Drug addicts (n= 17) and controls (n=20). >> >> 1.I ran a surface thickness study: >> >> ·For the ?mri_glmfit? command, I used DOSS and specified a contrast of >> +1 -1 0 (Controls > MA) >> >> ·I ran a pre-cached simulation with the following command: >> mri_glmfit-sim --glmdir lh.group_age.glmdir --cache 4 pos >> >> ·This simulation should only have corrected for the left hemisphere as >> I understand it. My cluster summary text file showed me that a cluster >> survived in the insula. I interpreted this as ?controls had higher >> grey matter thickness in this cluster located in the insula than drug >> addicts?. However, this was only corrected over the left hemisphere >> (when I ran the same simulation in the right hemisphere, nothing >> survived multiple comparison correction). >> >> ·I now wanted to correct the lh results for 2 hemispheres. From the >> help menu of ?mri_glmfit-sim?, I understood you can do this in 2 ways, >> are both of these correct and give the same result? >> >> a.mri_glmfit-sim --glmdir lh.group_age.glmdir --cache 4 pos --2 spaces >> --no-sim csdbase >> >> b.mri_glmfit-sim --glmdir lh.group_age.glmdir --cache 4 pos >> --cwpvalthresh 0.025 >> >> ØNo clusters survived the Bonferroni correction over both hemispheres. >> So I assume I can no longer report this result? >> >> 2.I ran a subcortical volume study: >> >> ·I used SPSS to perform the statistical analysis on each of 14 >> subcortical structures (left and right). Is there a good way to >> correct for multiple comparisons, apart from Bonferroni, which would >> be very conservative? >> >> Many thanks for your advice. >> >> Kind regards >> >> Reem >> >> *Reem Jan*** >> >> BPharm (Hons), RegPharmNZ >> >> PhD Student / Pharmacist >> >> School of Pharmacy, Faculty of Medical & Health Sciences, >> The University of Auckland, Private Bag 92019, >> Auckland, New Zealand. >> >> Ph: +64 9 373 7599 ext 81138. DDI: +64 9 923 1138 >> >> F: +64 9 367 7192 >> >> *From:*freesurfer-boun...@nmr.mgh.harvard.edu >> [mailto:freesurfer-boun...@nmr.mgh.harvard.edu] *On Behalf Of *Douglas >> Greve >> *Sent:* Friday, 20 July 2012 11:52 a.m. >> *To:* freesurfer@nmr.mgh.harvard.edu >> *Subject:* Re: [Freesurfer] slice-by-slice predictors >> >> Sorry, not possible. >> >> On 7/19/12 6:56 PM, Caspar M. Schwiedrzik wrote: >> >> Hi! >> Is it possible to have slice-by-slice predictors in Freesurfer, and if >> so, how? >> Thanks, >> Caspar >> >> >> >> >> _______________________________________________ >> Freesurfer mailing list >> Freesurfer@nmr.mgh.harvard.edu <mailto:Freesurfer@nmr.mgh.harvard.edu> >> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer >> >> >> _______________________________________________ >> Freesurfer mailing list >> Freesurfer@nmr.mgh.harvard.edu >> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer > > -- > Douglas N. Greve, Ph.D. > MGH-NMR Center > gr...@nmr.mgh.harvard.edu > Phone Number: 617-724-2358 > Fax: 617-726-7422 > > Bugs: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting > FileDrop: www.nmr.mgh.harvard.edu/facility/filedrop/index.html > > _______________________________________________ > Freesurfer mailing list > Freesurfer@nmr.mgh.harvard.edu > https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer > > > The information in this e-mail is intended only for the person to whom it is > addressed. If you believe this e-mail was sent to you in error and the e-mail > contains patient information, please contact the Partners Compliance HelpLine > at > http://www.partners.org/complianceline . If the e-mail was sent to you in > error > but does not contain patient information, please contact the sender and > properly > dispose of the e-mail. > > > _______________________________________________ > Freesurfer mailing list > Freesurfer@nmr.mgh.harvard.edu > https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer > > > _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
