Dear Manja,
It is statistically valid for the threshold suggested by FDR to vary in the different hemispheres. The reason is that FDR defines a proportion, not an absolute. The FDR procedure suggests a p-value threshold with the property that of all positive results at that threshold, one can expect that no more than X percent are false positives (where X is the desired FDR rate). It is very possible that the threshold with this property would be different in the right compared to the left hemisphere.
More informally, one is claiming that of all results declared "significant", X% are probably B.S. Since X is presumably small and the number of results large enough to be interesting, one can safely conclude that the experimental condition explains cortical variation at (most of) the locations indicated. Now divide the brain into subregions (perhaps right and left hemisphere, or perhaps something else) in each of which X% of the significant results are probably false. Then the claim that "X% of the overall results are probably false" is valid.
+glenn
------------------------------------ Here miracles become marvels, and marvels recurring wonders. -- William Beebe
Dr. Glenn Lawyer +352 661 967 244 Instituttgruppe for psykiatri Seksjon Vinderen PB 85 Vinderen 0319 Oslo http://folk.uio.no/davidgl
Dear Glen, there is something you have to take into account when using FDR separately at various regions of a parcelled brain. It do not control de Global FDR if the null hypothesis is true in many of these regions.
Although FDR provides strong control of false positives under the null hypothesis (similar to FWE) you have again a multiple comparison problem.
See Langers et al. / NeuroImage 38 (2007) 43–56
Cheers
Jorge
Phd Student Laboratory of functional Neuroscience. University of Pablo de Olavide. Seville. Spain
--- El jue, 28/8/08, d.g.lawyer@medisin.uio.no d.g.lawyer@medisin.uio.no escribió:
De: d.g.lawyer@medisin.uio.no d.g.lawyer@medisin.uio.no Asunto: [Freesurfer] Re: FDR correction Para: freesurfer@nmr.mgh.harvard.edu Fecha: jueves, 28 agosto, 2008 12:37 Dear Manja,
It is statistically valid for the threshold suggested by FDR to vary in the different hemispheres. The reason is that FDR defines a proportion, not an absolute. The FDR procedure suggests a p-value threshold with the property that of all positive results at that threshold, one can expect that no more than X percent are false positives (where X is the desired FDR rate). It is very possible that the threshold with this property would be different in the right compared to the left hemisphere.
More informally, one is claiming that of all results declared "significant", X% are probably B.S. Since X is presumably small and the number of results large enough to be interesting, one can safely conclude that the experimental condition explains cortical variation at (most of) the locations indicated. Now divide the brain into subregions (perhaps right and left hemisphere, or perhaps something else) in each of which X% of the significant results are probably false. Then the claim that "X% of the overall results are probably false" is valid.
+glenn
Here miracles become marvels, and marvels recurring wonders. -- William Beebe
Dr. Glenn Lawyer +352 661 967 244 Instituttgruppe for psykiatri Seksjon Vinderen PB 85 Vinderen 0319 Oslo http://folk.uio.no/davidgl
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
__________________________________________________ Correo Yahoo! Espacio para todos tus mensajes, antivirus y antispam ¡gratis! Regístrate ya - http://correo.yahoo.es
I think you are right
I bet you can choose to control FDR locally (clustering?)
2008/8/28 jorge luis jbernal0019@yahoo.es
Dear Glen, there is something you have to take into account when using FDR separately at various regions of a parcelled brain. It do not control de Global FDR if the null hypothesis is true in many of these regions.
Although FDR provides strong control of false positives under the null hypothesis (similar to FWE) you have again a multiple comparison problem.
See Langers et al. / NeuroImage 38 (2007) 43–56
Cheers
Jorge
Phd Student Laboratory of functional Neuroscience. University of Pablo de Olavide. Seville. Spain
--- El jue, 28/8/08, d.g.lawyer@medisin.uio.no d.g.lawyer@medisin.uio.no escribió:
De: d.g.lawyer@medisin.uio.no d.g.lawyer@medisin.uio.no Asunto: [Freesurfer] Re: FDR correction Para: freesurfer@nmr.mgh.harvard.edu Fecha: jueves, 28 agosto, 2008 12:37 Dear Manja,
It is statistically valid for the threshold suggested by FDR to vary in the different hemispheres. The reason is that FDR defines a proportion, not an absolute. The FDR procedure suggests a p-value threshold with the property that of all positive results at that threshold, one can expect that no more than X percent are false positives (where X is the desired FDR rate). It is very possible that the threshold with this property would be different in the right compared to the left hemisphere.
More informally, one is claiming that of all results declared "significant", X% are probably B.S. Since X is presumably small and the number of results large enough to be interesting, one can safely conclude that the experimental condition explains cortical variation at (most of) the locations indicated. Now divide the brain into subregions (perhaps right and left hemisphere, or perhaps something else) in each of which X% of the significant results are probably false. Then the claim that "X% of the overall results are probably false" is valid.
+glenn
Here miracles become marvels, and marvels recurring wonders. -- William Beebe
Dr. Glenn Lawyer +352 661 967 244 Instituttgruppe for psykiatri Seksjon Vinderen PB 85 Vinderen 0319 Oslo http://folk.uio.no/davidgl
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Correo Yahoo! Espacio para todos tus mensajes, antivirus y antispam ¡gratis! Regístrate ya - http://correo.yahoo.es
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
freesurfer@nmr.mgh.harvard.edu
-
d.g.lawyer＠medisin.uio.no -
jorge luis -
Pedro Paulo de Magalhães Oliveira Junior