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 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 <mailto:gr...@nmr.mgh.harvard.edu> >> To: freesurfer@nmr.mgh.harvard.edu >> <mailto: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 <mailto:gr...@nmr.mgh.harvard.edu> >> To: freesurfer@nmr.mgh.harvard.edu >> <mailto: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 >> <mailto:gr...@nmr.mgh.harvard.edu> >> To: freesurfer@nmr.mgh.harvard.edu >> <mailto: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 <mailto:r_maa...@hotmail.com> >> To: freesurfer@nmr.mgh.harvard.edu >> <mailto: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 >> <mailto:gr...@nmr.mgh.harvard.edu> >>> To: freesurfer@nmr.mgh.harvard.edu >> <mailto: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 >> <mailto:gr...@nmr.mgh.harvard.edu> >>>>> To: freesurfer@nmr.mgh.harvard.edu >> <mailto: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 >> <mailto: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 >> <mailto: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 >> <http://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 >> <mailto: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 >> <mailto: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 >> <mailto: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 >> <http://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 >> <mailto:Freesurfer@nmr.mgh.harvard.edu> >> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer >> >> _______________________________________________ >> Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu >> <mailto: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 >> <mailto:Freesurfer@nmr.mgh.harvard.edu> >> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer >> >> >> >> _______________________________________________ Freesurfer >> mailing list Freesurfer@nmr.mgh.harvard.edu >> <mailto: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 >> <mailto:Freesurfer@nmr.mgh.harvard.edu> >> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer >> >> >> >> _______________________________________________ Freesurfer >> mailing list Freesurfer@nmr.mgh.harvard.edu >> <mailto: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 >> <mailto:Freesurfer@nmr.mgh.harvard.edu> >> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer >> >> >> >> _______________________________________________ Freesurfer mailing >> list Freesurfer@nmr.mgh.harvard.edu >> <mailto: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 <mailto:Freesurfer@nmr.mgh.harvard.edu> >> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer >> >> >> >> _______________________________________________ Freesurfer mailing >> list Freesurfer@nmr.mgh.harvard.edu >> https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer 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 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 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
_______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
