[Freesurfer] design matrix palm
Hi all, I was adviced to use PALM for permutation of non-orthogonal surface data. However, I have some questions regarding the design matrix en contrast files to be used in PALM, but somehow I cannot subscribe to the FSL mailing list, and I was hoping one of the freesurfer users could help me out. I have a complex design and I would like to test for interactions as well for subgroups; in addition, I have several covariates. If I use the following collumns: A1 B1 A2 B2 cov1 cov2 cov3 I assumed I can at least test for the effects of A vs B (1 - 1 1 -1 0 0 0) and 1 vs 2 (1 1 -1 -1 0 0 0) and interactions (1 -1 -1 1 0 0 0). Q1: Is it true that I cannot compare subgroups, because columns denoted zero will be handeled as being a covariate? i.e. the contrast 1 -1 0 0 0 0 0 is inappropriate and I would have to build a new design matrix (A1 B1 cov1 cov2 cov3)? This is what I was told previously when using randomise for DTI data, but I'm not sure PALM works the same way. Q2: What I also understood from randomise is that you have to define both 1 -1 and -1 1; there is no option for a 2tailed t-test or F test. Is that correct and does it also apply to PALM? Thanks, Maaike ___ 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.
Re: [Freesurfer] mri_glmfit-sim permutation testing running after 3 days!
Hi all, Is using forced permutation for non-orthogonal design matrices wrong or is it allowed to do this instead of using tools like palm (what happens eg with the covariates when using forced permutation)? I used forced permutation and it seemed to work, results were (partly) comparable to what I found with monte carlo simulations. Thanks, Maaike Van: freesurfer-boun...@nmr.mgh.harvard.edunamens Harms, Michael Verzonden: vrijdag 26 augustus 2016 01:00:13 Aan: Freesurfer support list Onderwerp: Re: [Freesurfer] mri_glmfit-sim permutation testing running after 3 days! Hi, You might want to check out FSL’s PALM tool, which has a bit more sophisticated permutation framework, and allows for permutation in the context of non-orthogonal covariates. cheers, -MH -- Michael Harms, Ph.D. --- Conte Center for the Neuroscience of Mental Disorders Washington University School of Medicine Department of Psychiatry, Box 8134 660 South Euclid Ave. Tel: 314-747-6173 St. Louis, MO 63110 Email: mha...@wustl.edu From: > on behalf of Ajay Kurani > Reply-To: Freesurfer support list > Date: Thursday, August 25, 2016 at 4:13 PM To: Freesurfer support list > Subject: Re: [Freesurfer] mri_glmfit-sim permutation testing running after 3 days! Hi Doug, Thanks for the help! I think I figured out the issue based on your response. 1) I created a template to use for this group and named it fsaverage (including creating monte carlo simulations) for simplicity of integrating with freesurfer as I am newer to it. This is why the sizes didn't match up as you expected but the mri_glmfit still ran. 2) I deleted the folder and restarted without background processes. The error became apparent. Of my covariates (2 fix factors and 3 quantitative), not all were orthogonal. In looking at the error more, it seems that i need to add the --perm-force if I wanted the simulation to run, however the background processes were not aware of this error and kept polling as you mentioned. This brings me to a new but related issue. From what I have read in other freesurfer posts, it is statistically incorrect to use --perm-force for non-orthogonal covariates (or continuous covariates). I am unsure how to proceed. a) If I ran permutation testing (to overcome the issue of incorrect smoothness estimations from the gaussian distribution assumption), then I run into the issue of non-orthogonal covariates. Is there a way to orthogonalize the data in freesurfer, or a solution to this issue? b) If orthogonalizing is difficult to implement, another option is running Qdec with the montecarlo simulation at a more conservative p value (p< 0.001). From your previous posts, the testing at this p value for 10mm seems to meet the 5% FPR. One question is if the non-orthogonal data affects this analysis as well for this model? Thanks, Ajay On Thu, Aug 25, 2016 at 12:18 PM, Ajay Kurani > wrote: Hi Freesurfer Experts, I am trying to use freesurfer's mri_glmfit-sim tool to run permutation testing on cortical thickness data (as recommended by Doug in my previous post: http://www.mail-archive.com/freesurfer%40nmr.mgh.harvard.edu/msg48653.html ) Most of the tutorials I found were not related to permutation testing so the subsequent steps may be incorrect. Please let me know where I go wrong... 1) I first ran QDec to generate a folder for the analysis which would create the subsequent fsgd and y files needed my mri_glmfit-sim. I am running both left and right hemisphere cortical thickness analysis with 10mm smoothing. The following is for just the left hemisphere. Note I am doing a 3 group comparison, but for this 2 group ttest I manually centered the data based on the 3 group mean for age and education. 2) I ran the following command: /mri_glmfit-sim --glmdir ./HCvsPAT_lh_thickness_10mm/ --sim perm 1 2 perm.abs.2 --sim-sign abs --bg 16 Prior to running the command above, from the y.fsdg file I deleted the fwhm estimate of 13mm since this was not correctly estimated (ACF with long tails). I assumed that by removing this estimate, it would force the permutation test to calculate based on the data but when looking at the log output I see the following which says fwhm 0: cmdline mri_glmfit.bin --C ./HCvsPAT_lh_thickness_10mm//tmp.mri_glmfit-sim-19468/lh-Avg-Intercept-thickness.mtx --C ./HCvsPAT_lh_thickness_10mm//tmp.mri_glmfit-sim-19468/lh-Diff-Male-Female-Intercept-thickness.mtx --C
[Freesurfer] permutation/monte carlo
Dear Douglas, Sorry to bother you with this, but after reading the thread about the Eklund papers I'm getting confused about what I actually did when correcting for multiple comparisons and I'd also like to know whether I'm on the safe side. I used pre-cashed Monte Carlo simulation with a cluster finding threshold of 0.05 (at least so I thought): mri_glmfit -sim - -glmdir- -cache 1.3 abs - - cache-dir - - cwpvalthresh 0.01 Am I correct? And am I also correct that the FDR may be about 13% instead of 5% when doing this? (I smoothed thickness and surface area data with FWMH 10mm, but not the lGI data, because they were already very smooth and pre-cashed simulation didn't work after additional smoothing). Thanks, Maaike ___ 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.
Re: [Freesurfer] average values per cluster
Hi Douglas, Can I ask another question relating to an issue I ran into already a year ago? My lGI data were very smooth at the time (after smoothing for 10mm). I decided to use the unsmoothed data, since the fwmh is in the order of 10 then (according to the fwhm.dat file). My results are valid then, right? Just to check. For area and thickness I did smooth, resulting in comparable fwhm values. Thanks, Maaike Date: Mon, 9 Feb 2015 10:46:40 -0500 From: gr...@nmr.mgh.harvard.edu To: freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] average values per cluster On 2/8/15 1:48 PM, maaike rive wrote: But how can i run the appropriate preprocessing and smoothing? The --X flag s not recognised using mris_preproc. mri_preproc only uses the fsgd file to get the list of subjects to assure that they are ordered in the same way as when mri_glmfit builds the X matrix. I used the same matrix as X in an FSGD file to do this now, it seems to work, but I do not no if it is correct. Furthermore, If I try to run mri_glmfit --sim for the lGI data, I get an error because fwmh66 is not available. I did not smooth the lGI data and with the DODS/DOSS FSGD models there was no problem, so does this mean my new models are not correct? Wow, it is 66 without smoothing? I'm not sure what to tell you. The lGI is often very smooth, but that seems excessive. You can look for outliers by loading the y files as a "time coures" in tksurfer (-t flag). If it says at 66 then I don't think you can do the voxel-wise analysis. Was it that way when you used the FSGD file? It could be that your design matrix does not remove the mean offset, so check that too. > From: r_maa...@hotmail.com > Date: Fri, 6 Feb 2015 21:41:53 +0100 > To: freesurfer@nmr.mgh.harvard.edu > Subject: Re: [Freesurfer] average values per cluster > > Ok, thanks! > Maaike > > > Op 6 feb. 2015 om 17:49 heeft "Douglas N Greve" <gr...@nmr.mgh.harvard.edu> het volgende geschreven: > > > > > > There is not a way to build more flexible models in the FSGD structure. > > However, you can create your own design matrix and include anything you > > want in it and pass it to mri_glmfit with --X instead of --fsgd. > > > > doug > > > >> On 02/05/2015 06:13 AM, maaike rive wrote: > >> > >> Hi Doug, > >> > >> I hope I'm not driving you crazy, but I have some additional questions > >> regardig the DOSS/DODS models. Also because of the SPSS discrepancies. > >> > >> First, the model I discussed with you I used to assess diagn x age > >> interactions, regressing out the effects of state and gender. However, > >> in this model all interaction terms are incorporated. Is there a way > >> to build the model only incorporating main effects of diagn, state, > >> gender, age and diagn*age and diagn*state interactions? Since I do not > >> expect any higher order interactions or state* age interactions, for > >> instance (this is biologically not very plausible, although I did not > >> formally test it). > >> > >> Second, for areas where there is no diagn*age interaction I want to > >> use a model without any interaction term with age, so I was thinking > >> to use DOSS (I still want to regress out the effects of age, so it was > >> added as a regressor): > >> > >> diagn1*state1*gender1 > >> > >> diagn1*state1*gender2 > >> > >> diagn1*state2*gender1 > >> > >> diagn1*state2*gender2 > >> > >> diagn2*state1*gender1 > >> > >> diagn2*state1*gender2 > >> > >> diagn2*
[Freesurfer] visualization tk surfer
Dear Freesurfer experts, I have a question regarding tk surfer. If I load the cache.h13.abs.sig.cluster.mgh file, I'll only see the clusters surviving the statistical threshold I set, don't I? (e.g. I set the cluster finding threshold at 0.05 (Monte carlo simulation), keeping all clusters with p<0.001, which gives me for example one cluster; it's only this cluster that will be visualized, right?) I ask this because sometimes only a few clusters are kept, but I see a lot of the brain coloured in tk surfer, presumably because the clusters are very large. Thanks, Maaike___ 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.
Re: [Freesurfer] does lGI needs correction for any global measure?
Hi Doug, Thanks for the reply.I'm still struggling with this issue though. 1. For instance, if one corrects surface area for ICV in the GLM, shouldn't the fact that this relationship is non-linear be taken into account? Since volume increases with r^3 and surface with r^2... Same goes for lGI, how is the relationship between lGI en ICV? 2. If I choose to correct for TSA, it seems natural to correct thickness for mean (or total) thickness and lGI for some global measure (is there such thing as mean or total lGI?) as well, or am I mistaken? I don't understand why one measure should be corrected for a global measure with the same units, whereas another should not. 3. ICV is not influenced by the volumes of the different brain structures (it is not a summation of these volumes) so volume correction for ICV is unbiased. However TSA (or mean thickness, or "global lGI measure") is dependent of the area (or thickness or lGI) of each discrete brain structure, right? So for instance if one group has a smaller superior frontal area, total area will be smaller as well. Wouldn't correction remove part of the effect then? Maaike> To: freesurfer@nmr.mgh.harvard.edu; marie.sch...@unige.ch> From: gr...@nmr.mgh.harvard.edu> Date: Thu, 5 May 2016 12:32:11 -0400> Subject: Re: [Freesurfer] does lGI needs correction for any global measure?> > > > On 05/02/2016 07:39 AM, maaike rive wrote:> > Dear freesurfer experts,> > > >> >> > > > I have a question regarding correction of cortical measures or global > > measures (ICV, total surface area). As I understand from literature > > and the freesurfer wiki, a model for thickness should not be corrected > > for ICV (since thickness does not scale with total brain volume), but > > surface area should be corrected since is does scale with total brain > > volume. Three questions:> >> > > >> >> > > > 1. should I use ICV or should I use total surface area (TSA) as > > covariate in the model to correct surface area (by the way, average > > ICV and TSA do not differ between the groups to compare )?> >> >> Not sure. They will be different models. ICV should not change over the > lifetime whereas total surface area will.> > > > 2. does lGI also need correction? does it scale with volume? If so, do > > I need to correct for ICV or for TSA? Usually I see no correction but > > some do use ICV...> >> > > I don't know, perhaps Marie will weigh in> > > > 3. even if the average of certain measures (like ICV/TSA, or gender, > > or age) does not differ between groups, should they be added to the > > model? If I do not correct for gender, for example, my results change > > and most papers do use age and gender as covariates if they compare > > groups despite a lack of differences between groups...> >> >> They can still reduce intersubject variance.> doug> > > >> >> > > > Thanks,> >> > > >> >> > > > Maaike> >> >> >> >> > ___> > 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: https://gate.nmr.mgh.harvard.edu/filedrop2> www.nmr.mgh.harvard.edu/facility/filedrop/index.html> Outgoing: ftp://surfer.nmr.mgh.harvard.edu/transfer/outgoing/flat/greve/> > ___> 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 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] does lGI needs correction for any global measure?
Dear freesurfer experts,I have a question regarding correction of cortical measures or global measures (ICV, total surface area). As I understand from literature and the freesurfer wiki, a model for thickness should not be corrected for ICV (since thickness does not scale with total brain volume), but surface area should be corrected since is does scale with total brain volume. Three questions:1. should I use ICV or should I use total surface area (TSA) as covariate in the model to correct surface area (by the way, average ICV and TSA do not differ between the groups to compare )?2. does lGI also need correction? does it scale with volume? If so, do I need to correct for ICV or for TSA? Usually I see no correction but some do use ICV...3. even if the average of certain measures (like ICV/TSA, or gender, or age) does not differ between groups, should they be added to the model? If I do not correct for gender, for example, my results change and most papers do use age and gender as covariates if they compare groups despite a lack of differences between groups... Thanks,Maaike ___ 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.
Re: [Freesurfer] average values per cluster
But how can i run the appropriate preprocessing and smoothing? The --X flag s not recognised using mris_preproc. I used the same matrix as X in an FSGD file to do this now, it seems to work, but I do not no if it is correct. Furthermore, If I try to run mri_glmfit --sim for the lGI data, I get an error because fwmh66 is not available. I did not smooth the lGI data and with the DODS/DOSS FSGD models there was no problem, so does this mean my new models are not correct? From: r_maa...@hotmail.com Date: Fri, 6 Feb 2015 21:41:53 +0100 To: freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] average values per cluster Ok, thanks! Maaike Op 6 feb. 2015 om 17:49 heeft Douglas N Greve gr...@nmr.mgh.harvard.edu het volgende geschreven: There is not a way to build more flexible models in the FSGD structure. However, you can create your own design matrix and include anything you want in it and pass it to mri_glmfit with --X instead of --fsgd. doug On 02/05/2015 06:13 AM, maaike rive wrote: Hi Doug, I hope I'm not driving you crazy, but I have some additional questions regardig the DOSS/DODS models. Also because of the SPSS discrepancies. First, the model I discussed with you I used to assess diagn x age interactions, regressing out the effects of state and gender. However, in this model all interaction terms are incorporated. Is there a way to build the model only incorporating main effects of diagn, state, gender, age and diagn*age and diagn*state interactions? Since I do not expect any higher order interactions or state* age interactions, for instance (this is biologically not very plausible, although I did not formally test it). Second, for areas where there is no diagn*age interaction I want to use a model without any interaction term with age, so I was thinking to use DOSS (I still want to regress out the effects of age, so it was added as a regressor): diagn1*state1*gender1 diagn1*state1*gender2 diagn1*state2*gender1 diagn1*state2*gender2 diagn2*state1*gender1 diagn2*state1*gender2 diagn2*state2*gender1 diagn2*state2*gender2 age However, here there are still interaction terms with gender in the model. I do want to regress out the effects of gender (since I expect this to be a confounder), without incorporating the interaction term. How should I do this? The only solution I can think of is building a model with gender as additional regressor (containing 0 en 1) and using DOSS. So: diagn1*state1 diagn1*state2 diagn2*state1 diagn2*state2 age gender Does this make sense? Third, if so, it implies that for areas where there is no diagn*state interaction, and where I want to test diagn1 vs diagn2 (regressing out the effects of state, since I expect this to be a confounder), I should again build a new model: diagn1 diagn2 age gender state I realize this is a lot of work, hence I hope you could give me any advice about this. Thanks, Maaike From: r_maa...@hotmail.com To: freesurfer@nmr.mgh.harvard.edu Date: Tue, 3 Feb 2015 13:18:55 +0100 Subject: Re: [Freesurfer] average values per cluster Hi Doug, Sorry, the statistician doesn't understand it yet; we're currently building freesurfer and SPSS models step by step to find out what's going on, but it's a litte time consuming... so, to be continued! Maaike From: r_maa...@hotmail.com To: freesurfer@nmr.mgh.harvard.edu Date: Fri, 30 Jan 2015 20:55:27 +0100 Subject: Re: [Freesurfer] average values per cluster On second thought, I think the reason for the discrepancy is that I included state as a factor instead of covariate in the SPSS model. I ran the model again after having adjusted this and included every possible interaction; now I reach a p value of 0.009 (which is still different but at least significant). But I'll check again Monday! From: r_maa...@hotmail.com To: freesurfer@nmr.mgh.harvard.edu Date: Fri, 30 Jan 2015 20:15:07 +0100 Subject: Re: [Freesurfer] average values per cluster I thought so, but I will check with the statistician next Monday, I'll get back to you as soon as I'm sure. Date: Fri, 30 Jan 2015 14:12:32 -0500 From: gr...@nmr.mgh.harvard.edu To: freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] average values per cluster Actually, that looks correct. Is this what you fed to SPSS? On 1/30/15 2:07 PM, maaike rive wrote: Ok, I'm so sorry, I'm afraid I mixed things up. I indeed used
Re: [Freesurfer] average values per cluster
Ok, thanks! Maaike Op 6 feb. 2015 om 17:49 heeft Douglas N Greve gr...@nmr.mgh.harvard.edu het volgende geschreven: There is not a way to build more flexible models in the FSGD structure. However, you can create your own design matrix and include anything you want in it and pass it to mri_glmfit with --X instead of --fsgd. doug On 02/05/2015 06:13 AM, maaike rive wrote: Hi Doug, I hope I'm not driving you crazy, but I have some additional questions regardig the DOSS/DODS models. Also because of the SPSS discrepancies. First, the model I discussed with you I used to assess diagn x age interactions, regressing out the effects of state and gender. However, in this model all interaction terms are incorporated. Is there a way to build the model only incorporating main effects of diagn, state, gender, age and diagn*age and diagn*state interactions? Since I do not expect any higher order interactions or state* age interactions, for instance (this is biologically not very plausible, although I did not formally test it). Second, for areas where there is no diagn*age interaction I want to use a model without any interaction term with age, so I was thinking to use DOSS (I still want to regress out the effects of age, so it was added as a regressor): diagn1*state1*gender1 diagn1*state1*gender2 diagn1*state2*gender1 diagn1*state2*gender2 diagn2*state1*gender1 diagn2*state1*gender2 diagn2*state2*gender1 diagn2*state2*gender2 age However, here there are still interaction terms with gender in the model. I do want to regress out the effects of gender (since I expect this to be a confounder), without incorporating the interaction term. How should I do this? The only solution I can think of is building a model with gender as additional regressor (containing 0 en 1) and using DOSS. So: diagn1*state1 diagn1*state2 diagn2*state1 diagn2*state2 age gender Does this make sense? Third, if so, it implies that for areas where there is no diagn*state interaction, and where I want to test diagn1 vs diagn2 (regressing out the effects of state, since I expect this to be a confounder), I should again build a new model: diagn1 diagn2 age gender state I realize this is a lot of work, hence I hope you could give me any advice about this. Thanks, Maaike From: r_maa...@hotmail.com To: freesurfer@nmr.mgh.harvard.edu Date: Tue, 3 Feb 2015 13:18:55 +0100 Subject: Re: [Freesurfer] average values per cluster Hi Doug, Sorry, the statistician doesn't understand it yet; we're currently building freesurfer and SPSS models step by step to find out what's going on, but it's a litte time consuming... so, to be continued! Maaike From: r_maa...@hotmail.com To: freesurfer@nmr.mgh.harvard.edu Date: Fri, 30 Jan 2015 20:55:27 +0100 Subject: Re: [Freesurfer] average values per cluster On second thought, I think the reason for the discrepancy is that I included state as a factor instead of covariate in the SPSS model. I ran the model again after having adjusted this and included every possible interaction; now I reach a p value of 0.009 (which is still different but at least significant). But I'll check again Monday! From: r_maa...@hotmail.com To: freesurfer@nmr.mgh.harvard.edu Date: Fri, 30 Jan 2015 20:15:07 +0100 Subject: Re: [Freesurfer] average values per cluster I thought so, but I will check with the statistician next Monday, I'll get back to you as soon as I'm sure. Date: Fri, 30 Jan 2015 14:12:32 -0500 From: gr...@nmr.mgh.harvard.edu To: freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] average values per cluster Actually, that looks correct. Is this what you fed to SPSS? On 1/30/15 2:07 PM, maaike rive wrote: Ok, I'm so sorry, I'm afraid I mixed things up. I indeed used DODS, and assumed these were may regressors: diagn1*state1*gender1 diagn1*state1*gender2 diagn1*state2*gender1 diagn1*state2*gender2 diagn2*state1*gender1 diagn2*state1*gender2 diagn2*state2*gender1 diagn2*state2*gender2 diagn1*state1*gender1*age diagn1*state1*gender2*age diagn1*state2*gender1*age diagn1*state2*gender2*age diagn2*state1*gender1*age diagn2*state1*gender2*age diagn2*state2*gender1*age diagn2*state2*gender2*age However what I actually meant to do was testing for the effects of diagnosis and interactions (e.g. diagnosis*age) by using the appropriate contrasts, so I thought that if I used the contrast for the diagnosis*age interaction, I was testing
Re: [Freesurfer] defining ROIs/clusters
Won't I get the data for the complete eg insula then, instead of the cluster within the insula resulting from the comparison? Related to this: I noticed that in the aparc.stats files the magnitude of surface area is in the order of 100-1000 mm2, whereas in the ocn.dat files resulting from the clusterwise corrected analyses the average values of the resulting clusters are in the order of 0.something till 1 or 2, so a factor 1000 smaller. How can this be explained? Thanks again, Maaike Date: Tue, 3 Feb 2015 10:30:10 -0500From: gr...@nmr.mgh.harvard.eduTo: freesur...@nmr.mgh.harvard.eduSubject: Re: [Freesurfer] defining ROIs/clustersJust run mris_preproc on your healthy group, then run mri_segstats using the cluster annotation from your first analysis and the HC stack as the input. Does that make sense?doug On 2/3/15 9:18 AM, maaike rive wrote: Hi all, Does anyone know how to extract average (thickness/area/GI) values of a cluster A for subjects in a group which was NOT included in the model where this cluster A emerged? For example: in a t-test for diagnosis1dagnosis2 there is an effect in an insulacluster (found with mri_glmfit -sim); post hoc I want to test whether diagnosis1 and/or diagnosis2 differ from healthy controls in exactly the same cluster. I understand I can run the analyses in seperate models (one for diagnosis1 vs HC and one for diagnosis2 vs HC; if this cluster does not emerge in either model, there is no difference) but if I want to plot all groups for visualization purposes, I have no values for the HC. Thanks in advance, Maaike ___ 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 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 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.
Re: [Freesurfer] average values per cluster
Hi Doug, I hope I'm not driving you crazy, but I have some additional questions regardig the DOSS/DODS models. Also because of the SPSS discrepancies. First, the model I discussed with you I used to assess diagn x age interactions, regressing out the effects of state and gender. However, in this model all interaction terms are incorporated. Is there a way to build the model only incorporating main effects of diagn, state, gender, age and diagn*age and diagn*state interactions? Since I do not expect any higher order interactions or state* age interactions, for instance (this is biologically not very plausible, although I did not formally test it). Second, for areas where there is no diagn*age interaction I want to use a model without any interaction term with age, so I was thinking to use DOSS (I still want to regress out the effects of age, so it was added as a regressor): diagn1*state1*gender1 diagn1*state1*gender2 diagn1*state2*gender1 diagn1*state2*gender2 diagn2*state1*gender1 diagn2*state1*gender2 diagn2*state2*gender1 diagn2*state2*gender2 age However, here there are still interaction terms with gender in the model. I do want to regress out the effects of gender (since I expect this to be a confounder), without incorporating the interaction term. How should I do this? The only solution I can think of is building a model with gender as additional regressor (containing 0 en 1) and using DOSS. So: diagn1*state1 diagn1*state2 diagn2*state1 diagn2*state2 age gender Does this make sense? Third, if so, it implies that for areas where there is no diagn*state interaction, and where I want to test diagn1 vs diagn2 (regressing out the effects of state, since I expect this to be a confounder), I should again build a new model: diagn1 diagn2 age gender state I realize this is a lot of work, hence I hope you could give me any advice about this. Thanks, Maaike From: r_maaike@hotmail.comTo: freesur...@nmr.mgh.harvard.eduDate: Tue, 3 Feb 2015 13:18:55 +0100Subject: Re: [Freesurfer] average values per cluster Hi Doug, Sorry, the statistician doesn't understand it yet; we're currently building freesurfer and SPSS models step by step to find out what's going on, but it's a litte time consuming... so, to be continued! Maaike From: r_maaike@hotmail.comTo: freesur...@nmr.mgh.harvard.eduDate: Fri, 30 Jan 2015 20:55:27 +0100Subject: Re: [Freesurfer] average values per cluster On second thought, I think the reason for the discrepancy is that I included state as a factor instead of covariate in the SPSS model. I ran the model again after having adjusted this and included every possible interaction; now I reach a p value of 0.009 (which is still different but at least significant). But I'll check again Monday! From: r_maaike@hotmail.comTo: freesur...@nmr.mgh.harvard.eduDate: Fri, 30 Jan 2015 20:15:07 +0100Subject: Re: [Freesurfer] average values per cluster I thought so, but I will check with the statistician next Monday, I'll get back to you as soon as I'm sure. Date: Fri, 30 Jan 2015 14:12:32 -0500From: gr...@nmr.mgh.harvard.eduTo: freesur...@nmr.mgh.harvard.eduSubject: Re: [Freesurfer] average values per clusterActually, that looks correct. Is this what you fed to SPSS? On 1/30/15 2:07 PM, maaike rive wrote: Ok, I'm so sorry, I'm afraid I mixed things up. I indeed used DODS, and assumed these were may regressors: diagn1*state1*gender1 diagn1*state1*gender2 diagn1*state2*gender1 diagn1*state2*gender2 diagn2*state1*gender1 diagn2*state1*gender2 diagn2*state2*gender1 diagn2*state2*gender2 diagn1*state1*gender1*age diagn1*state1*gender2*age diagn1*state2*gender1*age diagn1*state2*gender2*age diagn2*state1*gender1*age diagn2*state1*gender2*age diagn2*state2*gender1*age diagn2*state2*gender2*age However what I actually meant to do was testing for the effects of diagnosis and interactions (e.g. diagnosis*age) by using the appropriate contrasts, so I thought that if I used the contrast for the diagnosis*age interaction, I was testing whether or not this interaction was significant regressing out the effects of state and gender. Another example, I thought that if I used the contrast for diagnosis1 versus diagnosis2, I was testing wether or not there was a significant effect of diagnosis regressing out the effects of state, gender and age. But that is not the case than? How should I build the model to answer these questions? Maaike Date: Fri, 30 Jan 2015 13:33:40 -0500From: gr...@nmr.mgh.harvard.eduTo: freesur...@nmr.mgh.harvard.eduSubject: Re: [Freesurfer] average values per clusterIt depends on the actual covariates and whether you used DOSS or DODS. The default is DODS, which is a full interaction model. Assuming that state is a categorical variable with 2 levels, you'd have 4 regressors (gender by state) + 4 more regressors (gender by state by age). If the SPSS model did not have 8 regressors
[Freesurfer] defining ROIs/clusters
Hi all, Does anyone know how to extract average (thickness/area/GI) values of a cluster A for subjects in a group which was NOT included in the model where this cluster A emerged? For example: in a t-test for diagnosis1dagnosis2 there is an effect in an insulacluster (found with mri_glmfit -sim); post hoc I want to test whether diagnosis1 and/or diagnosis2 differ from healthy controls in exactly the same cluster. I understand I can run the analyses in seperate models (one for diagnosis1 vs HC and one for diagnosis2 vs HC; if this cluster does not emerge in either model, there is no difference) but if I want to plot all groups for visualization purposes, I have no values for the HC. Thanks in advance, Maaike ___ 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.
Re: [Freesurfer] average values per cluster
Hi Doug, Sorry, the statistician doesn't understand it yet; we're currently building freesurfer and SPSS models step by step to find out what's going on, but it's a litte time consuming... so, to be continued! Maaike From: r_maaike@hotmail.comTo: freesur...@nmr.mgh.harvard.eduDate: Fri, 30 Jan 2015 20:55:27 +0100Subject: Re: [Freesurfer] average values per cluster On second thought, I think the reason for the discrepancy is that I included state as a factor instead of covariate in the SPSS model. I ran the model again after having adjusted this and included every possible interaction; now I reach a p value of 0.009 (which is still different but at least significant). But I'll check again Monday! From: r_maaike@hotmail.comTo: freesur...@nmr.mgh.harvard.eduDate: Fri, 30 Jan 2015 20:15:07 +0100Subject: Re: [Freesurfer] average values per cluster I thought so, but I will check with the statistician next Monday, I'll get back to you as soon as I'm sure. Date: Fri, 30 Jan 2015 14:12:32 -0500From: gr...@nmr.mgh.harvard.eduTo: freesur...@nmr.mgh.harvard.eduSubject: Re: [Freesurfer] average values per clusterActually, that looks correct. Is this what you fed to SPSS? On 1/30/15 2:07 PM, maaike rive wrote: Ok, I'm so sorry, I'm afraid I mixed things up. I indeed used DODS, and assumed these were may regressors: diagn1*state1*gender1 diagn1*state1*gender2 diagn1*state2*gender1 diagn1*state2*gender2 diagn2*state1*gender1 diagn2*state1*gender2 diagn2*state2*gender1 diagn2*state2*gender2 diagn1*state1*gender1*age diagn1*state1*gender2*age diagn1*state2*gender1*age diagn1*state2*gender2*age diagn2*state1*gender1*age diagn2*state1*gender2*age diagn2*state2*gender1*age diagn2*state2*gender2*age However what I actually meant to do was testing for the effects of diagnosis and interactions (e.g. diagnosis*age) by using the appropriate contrasts, so I thought that if I used the contrast for the diagnosis*age interaction, I was testing whether or not this interaction was significant regressing out the effects of state and gender. Another example, I thought that if I used the contrast for diagnosis1 versus diagnosis2, I was testing wether or not there was a significant effect of diagnosis regressing out the effects of state, gender and age. But that is not the case than? How should I build the model to answer these questions? Maaike Date: Fri, 30 Jan 2015 13:33:40 -0500From: gr...@nmr.mgh.harvard.eduTo: freesur...@nmr.mgh.harvard.eduSubject: Re: [Freesurfer] average values per clusterIt depends on the actual covariates and whether you used DOSS or DODS. The default is DODS, which is a full interaction model. Assuming that state is a categorical variable with 2 levels, you'd have 4 regressors (gender by state) + 4 more regressors (gender by state by age). If the SPSS model did not have 8 regressors, then it is not the same.doug On 1/30/15 1:26 PM, maaike rive wrote: How should I have described the model? Date: Fri, 30 Jan 2015 13:09:40 -0500From: gr...@nmr.mgh.harvard.eduTo: freesur...@nmr.mgh.harvard.eduSubject: Re: [Freesurfer] average values per clusterI told him it the model is a GLM with age, gender and state as covariatesThis is not a sufficient description. Several design matrices could be derived from this description. On 1/30/15 11:58 AM, maaike rive wrote: The p value from SPSS is 0.156. I checked with a statistician, just to be sure, but he says the SPSS model is correct... (he doesn't know anything about freesurfer and FSGD files, but I told him it the model is a GLM with age, gender and state as covariates). Date: Fri, 30 Jan 2015 11:46:37 -0500From: gr...@nmr.mgh.harvard.eduTo: freesur...@nmr.mgh.harvard.eduSubject: Re: [Freesurfer] average values per clusterYou should report the clusterwise p-value. The p-value that you have computed is not interpretable. If it did not come out significant, then it would be worrisome, but it is only a check and cannot be used for anything. What p-value did you get from SPSS? I'd be curious to know what the source of the descrpancy is since this has happened several times now.doug On 1/30/15 3:25 AM, maaike rive wrote: Hi Douglas, I used the matlab code and the p value is signifiant, although less so than the p-value I get from the clusterwise statistics (0.0027 vs 0.00010). (Indeed I used an abs threshold for the clusterwise statistics). Which p-value should I report? Thanks, Maaike From: r_maaike@hotmail.comTo: freesur...@nmr.mgh.harvard.eduDate: Thu, 29 Jan 2015 19:48:44 +0100Subject: Re: [Freesurfer] average values per cluster No not yet; I will do so! Date: Thu, 29 Jan 2015 13:44:33 -0500 From: gr...@nmr.mgh.harvard.edu To: freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] average values per cluster The t is correct. When the contrast matrix only has one row, then the F is just an unsigned t. I don't know anything about SPSS so I can't tell from
Re: [Freesurfer] average values per cluster
The p value from SPSS is 0.156. I checked with a statistician, just to be sure, but he says the SPSS model is correct... (he doesn't know anything about freesurfer and FSGD files, but I told him it the model is a GLM with age, gender and state as covariates). Date: Fri, 30 Jan 2015 11:46:37 -0500 From: gr...@nmr.mgh.harvard.edu To: freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] average values per cluster You should report the clusterwise p-value. The p-value that you have computed is not interpretable. If it did not come out significant, then it would be worrisome, but it is only a check and cannot be used for anything. What p-value did you get from SPSS? I'd be curious to know what the source of the descrpancy is since this has happened several times now. doug On 1/30/15 3:25 AM, maaike rive wrote: Hi Douglas, I used the matlab code and the p value is signifiant, although less so than the p-value I get from the clusterwise statistics (0.0027 vs 0.00010). (Indeed I used an abs threshold for the clusterwise statistics). Which p-value should I report? Thanks, Maaike From: r_maa...@hotmail.com To: freesurfer@nmr.mgh.harvard.edu Date: Thu, 29 Jan 2015 19:48:44 +0100 Subject: Re: [Freesurfer] average values per cluster No not yet; I will do so! Date: Thu, 29 Jan 2015 13:44:33 -0500 From: gr...@nmr.mgh.harvard.edu To: freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] average values per cluster The t is correct. When the contrast matrix only has one row, then the F is just an unsigned t. I don't know anything about SPSS so I can't tell from what you have sent whether it is the same model or not. Try to get SPSS to output the design matrix. Did you try the matlab code below? On 01/29/2015 01:40 PM, maaike rive wrote: Hi Douglas, I checked but as far as I see it I used the same models. I attached the SPSS model and output as wel as the FSGD file and contrast file (to test a diagnosis x age interaction). Now that I come to think of it, I think the reason for the discrepant findings is that I used a t-contrast instead of an F-contrast for the interaction in Freesurfer. Apologies I did not think about this earlier. But maybe there is something else I did completely wrong. What do you think? If it is indeed the t vs F contrast, than how should I specify that the contrast is an F-contrast? Is it ok to just add another line to the contrastfile (so 0 0 0 0 0 0 0 0 1 1 1 1 -1 -1 -1 -1 0 0 0 0 0 0 0 0 -1 -1 -1 -1 1 1 1 1)? Thanks again, Maaike Date: Mon, 26 Jan 2015 11:23:59 -0500 From: gr...@nmr.mgh.harvard.edu To: freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] average values per cluster We get these kind of reports occasionally. When I ask people to confirm that they use exactly the same design matrix in SPSS, I never hear back, so I assume that it gets resolved. So please check. The other thing you can do is to run in matlab, something like cd glmdir/contrast X = load('Xg.dat'); y = load('ocn.dat'); C = load('C.dat'); [beta rvar] = fast_glmfit(y,X); [F p] = fast_fratio(beta,X,rvar,C); p will be the p-value
Re: [Freesurfer] average values per cluster
I thought so, but I will check with the statistician next Monday, I'll get back to you as soon as I'm sure. Date: Fri, 30 Jan 2015 14:12:32 -0500 From: gr...@nmr.mgh.harvard.edu To: freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] average values per cluster Actually, that looks correct. Is this what you fed to SPSS? On 1/30/15 2:07 PM, maaike rive wrote: Ok, I'm so sorry, I'm afraid I mixed things up. I indeed used DODS, and assumed these were may regressors: diagn1*state1*gender1 diagn1*state1*gender2 diagn1*state2*gender1 diagn1*state2*gender2 diagn2*state1*gender1 diagn2*state1*gender2 diagn2*state2*gender1 diagn2*state2*gender2 diagn1*state1*gender1*age diagn1*state1*gender2*age diagn1*state2*gender1*age diagn1*state2*gender2*age diagn2*state1*gender1*age diagn2*state1*gender2*age diagn2*state2*gender1*age diagn2*state2*gender2*age However what I actually meant to do was testing for the effects of diagnosis and interactions (e.g. diagnosis*age) by using the appropriate contrasts, so I thought that if I used the contrast for the diagnosis*age interaction, I was testing whether or not this interaction was significant regressing out the effects of state and gender. Another example, I thought that if I used the contrast for diagnosis1 versus diagnosis2, I was testing wether or not there was a significant effect of diagnosis regressing out the effects of state, gender and age. But that is not the case than? How should I build the model to answer these questions? Maaike Date: Fri, 30 Jan 2015 13:33:40 -0500 From: gr...@nmr.mgh.harvard.edu To: freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] average values per cluster It depends on the actual covariates and whether you used DOSS or DODS. The default is DODS, which is a full interaction model. Assuming that state is a categorical variable with 2 levels, you'd have 4 regressors (gender by state) + 4 more regressors (gender by state by age). If the SPSS model did not have 8 regressors, then it is not the same. doug On 1/30/15 1:26 PM, maaike rive wrote: How should I have described the model? Date: Fri, 30 Jan 2015 13:09:40 -0500 From: gr...@nmr.mgh.harvard.edu To: freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] average values per cluster I told him it the model is a GLM with age, gender and state as covariates This is not a sufficient description. Several design matrices could be derived from this description. On 1/30/15 11:58 AM, maaike rive wrote: The p value from SPSS is 0.156. I checked with a statistician, just to be sure, but he says the SPSS model is correct... (he doesn't know anything about freesurfer and FSGD files, but I told him it the model is a GLM with age, gender and state as covariates). Date: Fri, 30 Jan 2015 11:46:37 -0500 From: gr...@nmr.mgh.harvard.edu To: freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] average values per cluster You should report the clusterwise p-value. The p-value that you have computed is not interpretable
Re: [Freesurfer] average values per cluster
Ok, I'm so sorry, I'm afraid I mixed things up. I indeed used DODS, and assumed these were may regressors: diagn1*state1*gender1 diagn1*state1*gender2 diagn1*state2*gender1 diagn1*state2*gender2 diagn2*state1*gender1 diagn2*state1*gender2 diagn2*state2*gender1 diagn2*state2*gender2 diagn1*state1*gender1*agediagn1*state1*gender2*age diagn1*state2*gender1*agediagn1*state2*gender2*agediagn2*state1*gender1*agediagn2*state1*gender2*agediagn2*state2*gender1*agediagn2*state2*gender2*age However what I actually meant to do was testing for the effects of diagnosis and interactions (e.g. diagnosis*age) by using the appropriate contrasts, so I thought that if I used the contrast for the diagnosis*age interaction, I was testing whether or not this interaction was significant regressing out the effects of state and gender. Another example, I thought that if I used the contrast for diagnosis1 versus diagnosis2, I was testing wether or not there was a significant effect of diagnosis regressing out the effects of state, gender and age. But that is not the case than? How should I build the model to answer these questions? Maaike Date: Fri, 30 Jan 2015 13:33:40 -0500 From: gr...@nmr.mgh.harvard.edu To: freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] average values per cluster It depends on the actual covariates and whether you used DOSS or DODS. The default is DODS, which is a full interaction model. Assuming that state is a categorical variable with 2 levels, you'd have 4 regressors (gender by state) + 4 more regressors (gender by state by age). If the SPSS model did not have 8 regressors, then it is not the same. doug On 1/30/15 1:26 PM, maaike rive wrote: How should I have described the model? Date: Fri, 30 Jan 2015 13:09:40 -0500 From: gr...@nmr.mgh.harvard.edu To: freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] average values per cluster I told him it the model is a GLM with age, gender and state as covariates This is not a sufficient description. Several design matrices could be derived from this description. On 1/30/15 11:58 AM, maaike rive wrote: The p value from SPSS is 0.156. I checked with a statistician, just to be sure, but he says the SPSS model is correct... (he doesn't know anything about freesurfer and FSGD files, but I told him it the model is a GLM with age, gender and state as covariates). Date: Fri, 30 Jan 2015 11:46:37 -0500 From: gr...@nmr.mgh.harvard.edu To: freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] average values per cluster You should report the clusterwise p-value. The p-value that you have computed is not interpretable. If it did not come out significant, then it would be worrisome, but it is only a check and cannot be used for anything. What p-value did you get from SPSS? I'd be curious to know what the source of the descrpancy is since this has happened several times now. doug On 1/30/15 3:25 AM, maaike rive wrote: Hi Douglas, I used the matlab code and the p value is signifiant, although less so than the p-value I get from the clusterwise statistics (0.0027 vs 0.00010). (Indeed I used an abs threshold for the clusterwise statistics). Which p-value should I report? Thanks, Maaike From: r_maa...@hotmail.com To: freesurfer@nmr.mgh.harvard.edu Date: Thu, 29 Jan 2015 19:48:44 +0100 Subject: Re: [Freesurfer] average values per cluster No not yet; I will do so! Date: Thu, 29 Jan 2015 13:44:33 -0500 From: gr...@nmr.mgh.harvard.edu To: freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] average values per cluster The t is correct. When the contrast matrix only has one row, then the F is just an unsigned t. I don't know anything about SPSS so I can't tell from what you have sent whether it is the same model or not. Try to get SPSS to output the design matrix. Did you try the matlab code below? On 01/29/2015 01:40 PM, maaike rive wrote: Hi Douglas, I checked but as far as I see it I used the same models. I attached the SPSS model and output as wel as the FSGD file and contrast file (to test a diagnosis x age interaction). Now that I come to think of it, I think the reason for the discrepant findings is that I used a t-contrast instead of an F-contrast for the interaction in Freesurfer. Apologies I did not think about this earlier. But maybe there is something else I did completely wrong. What do you think? If it is indeed the t vs F contrast, than how should I specify that the contrast is an F-contrast? Is it ok to just add another line to the contrastfile (so 0 0 0 0 0 0 0 0 1 1 1 1 -1 -1 -1 -1 0 0 0 0 0 0 0 0 -1 -1 -1 -1 1 1 1 1)? Thanks again, Maaike Date: Mon, 26 Jan 2015 11:23:59 -0500 From: gr...@nmr.mgh.harvard.edu To: freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] average values per cluster
Re: [Freesurfer] average values per cluster
How should I have described the model? Date: Fri, 30 Jan 2015 13:09:40 -0500 From: gr...@nmr.mgh.harvard.edu To: freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] average values per cluster I told him it the model is a GLM with age, gender and state as covariates This is not a sufficient description. Several design matrices could be derived from this description. On 1/30/15 11:58 AM, maaike rive wrote: The p value from SPSS is 0.156. I checked with a statistician, just to be sure, but he says the SPSS model is correct... (he doesn't know anything about freesurfer and FSGD files, but I told him it the model is a GLM with age, gender and state as covariates). Date: Fri, 30 Jan 2015 11:46:37 -0500 From: gr...@nmr.mgh.harvard.edu To: freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] average values per cluster You should report the clusterwise p-value. The p-value that you have computed is not interpretable. If it did not come out significant, then it would be worrisome, but it is only a check and cannot be used for anything. What p-value did you get from SPSS? I'd be curious to know what the source of the descrpancy is since this has happened several times now. doug On 1/30/15 3:25 AM, maaike rive wrote: Hi Douglas, I used the matlab code and the p value is signifiant, although less so than the p-value I get from the clusterwise statistics (0.0027 vs 0.00010). (Indeed I used an abs threshold for the clusterwise statistics). Which p-value should I report? Thanks, Maaike From: r_maa...@hotmail.com To: freesurfer@nmr.mgh.harvard.edu Date: Thu, 29 Jan 2015 19:48:44 +0100 Subject: Re: [Freesurfer] average values per cluster No not yet; I will do so! Date: Thu, 29 Jan 2015 13:44:33 -0500 From: gr...@nmr.mgh.harvard.edu To: freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] average values per cluster The t is correct. When the contrast matrix only has one row, then the F is just an unsigned t. I don't know anything about SPSS so I can't tell from what you have sent whether it is the same model or not. Try to get SPSS to output the design matrix. Did you try the matlab code below? On 01/29/2015 01:40 PM, maaike rive wrote: Hi Douglas, I checked but as far as I see it I used the same models. I attached the SPSS model and output as wel as the FSGD file and contrast file (to test a diagnosis x age interaction). Now that I come to think of it, I think the reason for the discrepant findings is that I used a t-contrast instead of an F-contrast for the interaction in Freesurfer. Apologies I did not think about this earlier. But maybe there is something else I did completely wrong. What do you think? If it is indeed the t vs F contrast, than how should I specify that the contrast is an F-contrast? Is it ok to just add another line to the contrastfile (so 0 0 0 0 0 0 0 0 1 1 1 1 -1 -1 -1 -1 0 0 0 0 0 0 0 0 -1 -1 -1 -1 1 1 1 1)? Thanks again, Maaike
Re: [Freesurfer] average values per cluster
On second thought, I think the reason for the discrepancy is that I included state as a factor instead of covariate in the SPSS model. I ran the model again after having adjusted this and included every possible interaction; now I reach a p value of 0.009 (which is still different but at least significant). But I'll check again Monday! From: r_maa...@hotmail.com To: freesurfer@nmr.mgh.harvard.edu Date: Fri, 30 Jan 2015 20:15:07 +0100 Subject: Re: [Freesurfer] average values per cluster I thought so, but I will check with the statistician next Monday, I'll get back to you as soon as I'm sure. Date: Fri, 30 Jan 2015 14:12:32 -0500 From: gr...@nmr.mgh.harvard.edu To: freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] average values per cluster Actually, that looks correct. Is this what you fed to SPSS? On 1/30/15 2:07 PM, maaike rive wrote: Ok, I'm so sorry, I'm afraid I mixed things up. I indeed used DODS, and assumed these were may regressors: diagn1*state1*gender1 diagn1*state1*gender2 diagn1*state2*gender1 diagn1*state2*gender2 diagn2*state1*gender1 diagn2*state1*gender2 diagn2*state2*gender1 diagn2*state2*gender2 diagn1*state1*gender1*age diagn1*state1*gender2*age diagn1*state2*gender1*age diagn1*state2*gender2*age diagn2*state1*gender1*age diagn2*state1*gender2*age diagn2*state2*gender1*age diagn2*state2*gender2*age However what I actually meant to do was testing for the effects of diagnosis and interactions (e.g. diagnosis*age) by using the appropriate contrasts, so I thought that if I used the contrast for the diagnosis*age interaction, I was testing whether or not this interaction was significant regressing out the effects of state and gender. Another example, I thought that if I used the contrast for diagnosis1 versus diagnosis2, I was testing wether or not there was a significant effect of diagnosis regressing out the effects of state, gender and age. But that is not the case than? How should I build the model to answer these questions? Maaike Date: Fri, 30 Jan 2015 13:33:40 -0500 From: gr...@nmr.mgh.harvard.edu To: freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] average values per cluster It depends on the actual covariates and whether you used DOSS or DODS. The default is DODS, which is a full interaction model. Assuming that state is a categorical variable with 2 levels, you'd have 4 regressors (gender by state) + 4 more regressors (gender by state by age). If the SPSS model did not have 8 regressors, then it is not the same. doug On 1/30/15 1:26 PM, maaike rive wrote: How should I have described the model? Date: Fri, 30 Jan 2015 13:09:40 -0500 From: gr...@nmr.mgh.harvard.edu To: freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] average values per cluster I told him it the model is a GLM with age, gender and state as covariates This is not a sufficient description. Several design matrices could be derived from this description. On 1/30/15 11:58 AM, maaike rive wrote: The p value from SPSS is 0.156. I checked with a statistician, just to be sure, but he says the SPSS model is correct... (he doesn't know anything about freesurfer and FSGD files, but I told him it the model is a GLM with age, gender and state as covariates). Date: Fri, 30 Jan 2015
Re: [Freesurfer] average values per cluster
Hi Douglas, I used the matlab code and the p value is signifiant, although less so than the p-value I get from the clusterwise statistics (0.0027 vs 0.00010). (Indeed I used an abs threshold for the clusterwise statistics). Which p-value should I report? Thanks, Maaike From: r_maa...@hotmail.com To: freesurfer@nmr.mgh.harvard.edu Date: Thu, 29 Jan 2015 19:48:44 +0100 Subject: Re: [Freesurfer] average values per cluster No not yet; I will do so! Date: Thu, 29 Jan 2015 13:44:33 -0500 From: gr...@nmr.mgh.harvard.edu To: freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] average values per cluster The t is correct. When the contrast matrix only has one row, then the F is just an unsigned t. I don't know anything about SPSS so I can't tell from what you have sent whether it is the same model or not. Try to get SPSS to output the design matrix. Did you try the matlab code below? On 01/29/2015 01:40 PM, maaike rive wrote: Hi Douglas, I checked but as far as I see it I used the same models. I attached the SPSS model and output as wel as the FSGD file and contrast file (to test a diagnosis x age interaction). Now that I come to think of it, I think the reason for the discrepant findings is that I used a t-contrast instead of an F-contrast for the interaction in Freesurfer. Apologies I did not think about this earlier. But maybe there is something else I did completely wrong. What do you think? If it is indeed the t vs F contrast, than how should I specify that the contrast is an F-contrast? Is it ok to just add another line to the contrastfile (so 0 0 0 0 0 0 0 0 1 1 1 1 -1 -1 -1 -1 0 0 0 0 0 0 0 0 -1 -1 -1 -1 1 1 1 1)? Thanks again, Maaike Date: Mon, 26 Jan 2015 11:23:59 -0500 From: gr...@nmr.mgh.harvard.edu To: freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] average values per cluster We get these kind of reports occasionally. When I ask people to confirm that they use exactly the same design matrix in SPSS, I never hear back, so I assume that it gets resolved. So please check. The other thing you can do is to run in matlab, something like cd glmdir/contrast X = load('Xg.dat'); y = load('ocn.dat'); C = load('C.dat'); [beta rvar] = fast_glmfit(y,X); [F p] = fast_fratio(beta,X,rvar,C); p will be the p-value If you used an unsigned cluster-forming threshold (ie, abs), then it is possible that some of the voxels are pos and some are neg so that they average out doug On 01/26/2015 09:03 AM, maaike rive wrote: Dear Freesurfer experts, Sorry to bother you again, but I have two more questions about extracting (thickness/surface/GI) values from a certain cluster. As I understood, the abs.y.ocn.dat file gives the average values for a given significant cluster (e.g. a cluster where there is a significant AxB interaction). I may be completely misunderstanding things, but if I use these values in SPSS for further statistics and test the same interaction (AxB), than according to SPSS this interaction is /not /significant (corrected for the same covariates as in the FSGD file). Could you tell me what is going wrong here? I do not trust my results now. Furthermore, is it possible (and if so, how?) to extract the average values of exactly the same cluster, but in an independent group not used in the analysis, for post-hoc comparisons in SPSS? Thanks, Maaike ___ 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: https://gate.nmr.mgh.harvard.edu/filedrop2 www.nmr.mgh.harvard.edu/facility/filedrop/index.html Outgoing: ftp://surfer.nmr.mgh.harvard.edu/transfer/outgoing/flat/greve/ ___ 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 -- Douglas N. Greve
Re: [Freesurfer] average values per cluster
Hi Douglas, I checked but as far as I see it I used the same models. I attached the SPSS model and output as wel as the FSGD file and contrast file (to test a diagnosis x age interaction). Now that I come to think of it, I think the reason for the discrepant findings is that I used a t-contrast instead of an F-contrast for the interaction in Freesurfer. Apologies I did not think about this earlier. But maybe there is something else I did completely wrong. What do you think? If it is indeed the t vs F contrast, than how should I specify that the contrast is an F-contrast? Is it ok to just add another line to the contrastfile (so0 0 0 0 0 0 0 0 1 1 1 1 -1 -1 -1 -10 0 0 0 0 0 0 0 -1 -1 -1 -1 1 1 1 1)? Thanks again, Maaike Date: Mon, 26 Jan 2015 11:23:59 -0500 From: gr...@nmr.mgh.harvard.edu To: freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] average values per cluster We get these kind of reports occasionally. When I ask people to confirm that they use exactly the same design matrix in SPSS, I never hear back, so I assume that it gets resolved. So please check. The other thing you can do is to run in matlab, something like cd glmdir/contrast X = load('Xg.dat'); y = load('ocn.dat'); C = load('C.dat'); [beta rvar] = fast_glmfit(y,X); [F p] = fast_fratio(beta,X,rvar,C); p will be the p-value If you used an unsigned cluster-forming threshold (ie, abs), then it is possible that some of the voxels are pos and some are neg so that they average out doug On 01/26/2015 09:03 AM, maaike rive wrote: Dear Freesurfer experts, Sorry to bother you again, but I have two more questions about extracting (thickness/surface/GI) values from a certain cluster. As I understood, the abs.y.ocn.dat file gives the average values for a given significant cluster (e.g. a cluster where there is a significant AxB interaction). I may be completely misunderstanding things, but if I use these values in SPSS for further statistics and test the same interaction (AxB), than according to SPSS this interaction is /not /significant (corrected for the same covariates as in the FSGD file). Could you tell me what is going wrong here? I do not trust my results now. Furthermore, is it possible (and if so, how?) to extract the average values of exactly the same cluster, but in an independent group not used in the analysis, for post-hoc comparisons in SPSS? Thanks, Maaike ___ 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: https://gate.nmr.mgh.harvard.edu/filedrop2 www.nmr.mgh.harvard.edu/facility/filedrop/index.html Outgoing: ftp://surfer.nmr.mgh.harvard.edu/transfer/outgoing/flat/greve/ ___ 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. FSGD_diagnbystate_ex.txt~ copy Description: Binary data 0 0 0 0 0 0 0 0 1 1 1 1 -1 -1 -1 -1 SPSS model and output.docx Description: MS-Word 2007 document ___ 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.
Re: [Freesurfer] average values per cluster
No not yet; I will do so! Date: Thu, 29 Jan 2015 13:44:33 -0500 From: gr...@nmr.mgh.harvard.edu To: freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] average values per cluster The t is correct. When the contrast matrix only has one row, then the F is just an unsigned t. I don't know anything about SPSS so I can't tell from what you have sent whether it is the same model or not. Try to get SPSS to output the design matrix. Did you try the matlab code below? On 01/29/2015 01:40 PM, maaike rive wrote: Hi Douglas, I checked but as far as I see it I used the same models. I attached the SPSS model and output as wel as the FSGD file and contrast file (to test a diagnosis x age interaction). Now that I come to think of it, I think the reason for the discrepant findings is that I used a t-contrast instead of an F-contrast for the interaction in Freesurfer. Apologies I did not think about this earlier. But maybe there is something else I did completely wrong. What do you think? If it is indeed the t vs F contrast, than how should I specify that the contrast is an F-contrast? Is it ok to just add another line to the contrastfile (so 0 0 0 0 0 0 0 0 1 1 1 1 -1 -1 -1 -1 0 0 0 0 0 0 0 0 -1 -1 -1 -1 1 1 1 1)? Thanks again, Maaike Date: Mon, 26 Jan 2015 11:23:59 -0500 From: gr...@nmr.mgh.harvard.edu To: freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] average values per cluster We get these kind of reports occasionally. When I ask people to confirm that they use exactly the same design matrix in SPSS, I never hear back, so I assume that it gets resolved. So please check. The other thing you can do is to run in matlab, something like cd glmdir/contrast X = load('Xg.dat'); y = load('ocn.dat'); C = load('C.dat'); [beta rvar] = fast_glmfit(y,X); [F p] = fast_fratio(beta,X,rvar,C); p will be the p-value If you used an unsigned cluster-forming threshold (ie, abs), then it is possible that some of the voxels are pos and some are neg so that they average out doug On 01/26/2015 09:03 AM, maaike rive wrote: Dear Freesurfer experts, Sorry to bother you again, but I have two more questions about extracting (thickness/surface/GI) values from a certain cluster. As I understood, the abs.y.ocn.dat file gives the average values for a given significant cluster (e.g. a cluster where there is a significant AxB interaction). I may be completely misunderstanding things, but if I use these values in SPSS for further statistics and test the same interaction (AxB), than according to SPSS this interaction is /not /significant (corrected for the same covariates as in the FSGD file). Could you tell me what is going wrong here? I do not trust my results now. Furthermore, is it possible (and if so, how?) to extract the average values of exactly the same cluster, but in an independent group not used in the analysis, for post-hoc comparisons in SPSS? Thanks, Maaike ___ 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: https://gate.nmr.mgh.harvard.edu/filedrop2 www.nmr.mgh.harvard.edu/facility/filedrop/index.html Outgoing: ftp://surfer.nmr.mgh.harvard.edu/transfer/outgoing/flat/greve/ ___ 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 -- 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: https://gate.nmr.mgh.harvard.edu/filedrop2 www.nmr.mgh.harvard.edu/facility/filedrop/index.html Outgoing: ftp://surfer.nmr.mgh.harvard.edu/transfer/outgoing/flat/greve/ ___ Freesurfer mailing list Freesurfer
[Freesurfer] average values per cluster
Dear Freesurfer experts, Sorry to bother you again, but I have two more questions about extracting (thickness/surface/GI) values from a certain cluster. As I understood, the abs.y.ocn.dat file gives the average values for a given significant cluster (e.g. a cluster where there is a significant AxB interaction). I may be completely misunderstanding things, but if I use these values in SPSS for further statistics and test the same interaction (AxB), than according to SPSS this interaction is not significant (corrected for the same covariates as in the FSGD file). Could you tell me what is going wrong here? I do not trust my results now. Furthermore, is it possible (and if so, how?) to extract the average values of exactly the same cluster, but in an independent group not used in the analysis, for post-hoc comparisons in SPSS? Thanks, Maaike___ 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] Extracting vertexvalues
Dear Freesurfer experts, I have a question about extracting vertex values of VtxMax for each subject (for surface eara, in this case). Based on earlier postings, I used the command mri_segstats (--seg ${x}/surf/lh/area.pial --i ${x}/surf/lh/area.pial --vox vertexno 0 0). First, I'm not exactly sure whether this is correct and second, I get 2 columns whereas I expected 1 column, so I wonder, what do these two values per subject mean? Thanking you in advance, Maaike ___ 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.
Re: [Freesurfer] Extracting vertexvalues
Thanks! Date: Mon, 5 Jan 2015 14:57:12 -0500 From: gr...@nmr.mgh.harvard.edu To: freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] Extracting vertexvalues It is reporting both the value at that vertex and the mean of all the other vertices. To get only the value for the vertex, add --id 1 doug On 01/05/2015 03:52 AM, maaike rive wrote: Dear Freesurfer experts, I have a question about extracting vertex values of VtxMax for each subject (for surface eara, in this case). Based on earlier postings, I used the command mri_segstats (--seg ${x}/surf/lh/area.pial --i ${x}/surf/lh/area.pial --vox vertexno 0 0). First, I'm not exactly sure whether this is correct and second, I get 2 columns whereas I expected 1 column, so I wonder, what do these two values per subject mean? Thanking you in advance, Maaike ___ 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: https://gate.nmr.mgh.harvard.edu/filedrop2 www.nmr.mgh.harvard.edu/facility/filedrop/index.html Outgoing: ftp://surfer.nmr.mgh.harvard.edu/transfer/outgoing/flat/greve/ ___ 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 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] DODS design and demeaning
Dear Freesurfer experts, I would be very greatful if you could help me with three (probably basic) questions regarding an mri_glmfit design. For the design and contrasts I used FSGDF. I have a 2x2 factorial design with one covariate (age). I specified all four groups (A1B1, A1B2, A2B1, A2B2) and the covariate in one model, because I would like to test for the interaction between A and B (regressing out the effects of age). Age was demeaned using the mean of all subjects (i.e. of all 4 groups). My first question regards the contrasts: If, besides the interaction and main affects of A and B, I would like to assess the difference between individual groups, for example A1B1 - A1B2, is it okay to use the same model (1 -1 0 0 + further 0's for the age regressors); or should I make a new FSGDF containing only the subjects of group A1B1 and A1B2? My second question is associated with the first: If it is allowed to use the same model, how does freesurfer deal withdemeaning? Since I used the mean of all subjects to demean and not just of the subjects of for example A1B1 and A1B2. My third question regards the DODS/DOSS issue. I used DODS, because I do not believe the slopes of age by volume (or age by area/thickness) will be exactly parallel between groups. I tested this also be using contrasts of the age regressors, (for example using the contrast 0 0 0 0 1 -1 -1 1 for a AxBxage interaction) and indeed there are interactions with age in some brain areas. For thickness, they were FDR corrected still significant, for other measeurs the were not significant after correction, but that seems natural given the sample sizes. No I wondered whether it would suffice to use a DODS design (since slopes differ between groups) or whether I should stratify groups further for age (for example use two models, one for young subjects, on for old subjects). If the latter is the case, I would end up with very small samples, I am afraid. Thanking you in advance, Maaike___ 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.
Re: [Freesurfer] DODS design and demeaning
Hi Doug, Thank you for your quick response. Concerning question3, do you mean that if there is an interaction with age, it is all right to use DODS (without stratification)? Thanks, Maaike Date: Wed, 24 Dec 2014 15:31:53 -0500 From: gr...@nmr.mgh.harvard.edu To: freesurfer@nmr.mgh.harvard.edu Subject: Re: [Freesurfer] DODS design and demeaning On 12/24/14 5:21 AM, maaike rive wrote: Dear Freesurfer experts, I would be very greatful if you could help me with three (probably basic) questions regarding an mri_glmfit design. For the design and contrasts I used FSGDF. I have a 2x2 factorial design with one covariate (age). I specified all four groups (A1B1, A1B2, A2B1, A2B2) and the covariate in one model, because I would like to test for the interaction between A and B (regressing out the effects of age). Age was demeaned using the mean of all subjects (i.e. of all 4 groups). My first question regards the contrasts: If, besides the interaction and main affects of A and B, I would like to assess the difference between individual groups, for example A1B1 - A1B2, is it okay to use the same model (1 -1 0 0 + further 0's for the age regressors); or should I make a new FSGDF containing only the subjects of group A1B1 and A1B2? It is a tricky question. You will get the same group means regardless of whether you combine all subjects or use a 2nd FSGD. The difference will be that you will have a much higher DOF if you combine them all. This means that the noise in the other subjects informs you about the noise in those two groups. If you can justify this, then you can leave them combined. If the 2nd model gives you the results you want, then reporting that would be a stronger scientific result. My second question is associated with the first: If it is allowed to use the same model, how does freesurfer deal withdemeaning? Since I used the mean of all subjects to demean and not just of the subjects of for example A1B1 and A1B2. The demeaning is also tricky. I would actually test whether there is an interaction between group and age. If there is no interaction, then I would re-run using a DOSS model in which case the demeaning won't make a difference. My third question regards the DODS/DOSS issue. I used DODS, because I do not believe the slopes of age by volume (or age by area/thickness) will be exactly parallel between groups. I tested this also be using contrasts of the age regressors, (for example using the contrast 0 0 0 0 1 -1 -1 1 for a AxBxage interaction) and indeed there are interactions with age in some brain areas. For thickness, they were FDR corrected still significant, for other measeurs the were not significant after correction, but that seems natural given the sample sizes. No I wondered whether it would suffice to use a DODS design (since slopes differ between groups) or whether I should stratify groups further for age (for example use two models, one for young subjects, on for old subjects). If the latter is the case, I would end up with very small samples, I am afraid. If there is no interaction in the areas that show up in your contrast of interest, then it is safe to use DOSS. If there are, then I don't think that stratifying them solves the problem. doug Thanking you in advance, Maaike ___ 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 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 The information in this e-mail is intended only for the person to whom