External Email - Use Caution Dr. Greve,
Thank you for this information. The input to mri_glmfit was an mgh file of structural MRI data from several subjects warped into a common space and concatenated. What would that make the units of the input (--y) file and the output gamma file? To the second question, the analysis is continuous (association of several continuous cognitive variables and area/thickness/volume) so I am thinking perhaps the output file being produced by this operation is not Cohen's d. Is this possible or does that operation always produce Cohen's d? I found a line on the documentation that says - "For measuring the contrast to noise ratio (CNR) in mri_glmfit, the file cnr.nii is generated by the contrast value (gamma) divided by the residual stddev (rstd)." Does that make my output file a contrast to noise ratio? Max On Thu, Jan 10, 2019 at 11:40 AM Greve, Douglas N.,Ph.D. < dgr...@mgh.harvard.edu> wrote: > > > On 1/9/19 10:35 AM, Max Owens wrote: > > > > External Email - Use Caution > > > > Hi, > > > > I just wanted to follow up and see if there was anyone who could > > answer my questions: > > > > 1.What is the unit of effect size for the gamma.mgh file created by > > vertexwise GLM Analysis using mri_glmfit? > Same units as the input (ie, --y) > > > > 2.Based on two previous posts > > ( > https://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg52144.html and > > > > https://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg57316.html) > > > I tried to calculate Cohen’s d in all significant clusters resulting > > from clusterwise multiple comparison correction of a vertexwise GLM > > using mri_glmfit. I did this by dividing the gamma.mgh file by > > rstd.mgh to create a cohensd.mgh file (fscalc gamma.mgh div > > ../rstd.mgh -o cohensd.mgh). Then from the output of that operation I > > extracted the mean value in each significant cluster from my > > vertexwise analysis using mri_segstats (mri_segstats --i cohensd.mgh > > --seg cluster.sig.ocn.mgh --exclude 0 --o sum.dat). My understanding > > was that this would yield the average Cohen’s d of each cluster. > > However, if it is indeed Cohen’s d the effect sizes yielded (between > > .01 and .02) are so small that they should be undetectable given my > > sample size (N=1104). This leads me to believe that the outputs are > > not in fact the average Cohen’s d of the cluster. Can anyone provide > > any clarification of what the output file of this process means? > > Specifically, what is the unit of effect size? > Cohen's D is defined for a two group analysis as > (mean1-mean2)/PooledStdDev. Do you have a two-group design? If not, then > Cohen's D might not be the right effect size to use. > > > > Please let me know if I can clarify anything. > > > > Thanks, > > > > Max > > > > On Fri, Dec 7, 2018 at 1:02 PM Max Owens <owensma...@gmail.com > > <mailto:owensma...@gmail.com>> wrote: > > > > Hi Freesurfer Experts, > > > > I have two related questions about the measures of effect size > > that can be derived from a vertexwise GLM analyses conducted in > > Freesurfer: > > > > 1.What is the unit of effect size for the gamma.mgh file created > > by vertexwise GLM Analysis using mri_glmfit? > > > > 2.Based on two previous posts > > ( > https://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg52144.html > > and > > > https://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg57316.html) > > I tried to calculate Cohen’s d in all significant clusters > > resulting from clusterwise multiple comparison correction of a > > vertexwise GLM using mri_glmfit. I did this by dividing the > > gamma.mgh file by rstd.mgh to create a cohensd.mgh file (fscalc > > gamma.mgh div ../rstd.mgh -o cohensd.mgh). Then from the output of > > that operation I extracted the mean value in each significant > > cluster from my vertexwise analysis using mri_segstats > > (mri_segstats --i cohensd.mgh --seg cluster.sig.ocn.mgh --exclude > > 0 --o sum.dat). My understanding was that this would yield the > > average Cohen’s d of each cluster. However, if it is indeed > > Cohen’s d the effect sizes yielded (between .01 and .02) are so > > small that they should be undetectable given my sample size > > (N=1104). This leads me to believe that the outputs are not in > > fact the average Cohen’s d of the cluster. Can anyone provide any > > clarification of what the output file of this process means? > > Specifically, what is the unit of effect size? > > > > Thanks in advance for your help! > > > > Max > > > > > > > > -- > > > > Max M Owens > > > > Postdoctoral Fellow > > > > NERVE Laboratory > > > > University of Vermont > > > > > > _______________________________________________ > > 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 -- Max M Owens Postdoctoral Fellow NERVE Laboratory University of Vermont
_______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
