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, 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
cluster>>>I 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 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@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_______________________________________________
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer The information in
this e-mail is intended only for the person to whom it is addressed. If you
believe this e-mail was sent to you in error and the e-mail contains patient
information, please contact the Partners Compliance HelpLine at
http://www.partners.org/complianceline . If the e-mail was sent to you in error
but does not contain patient information, please contact the sender and
properly dispose of the e-mail.
_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer_______________________________________________
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer The information in
this e-mail is intended only for the person to whom it is addressed. If you
believe this e-mail was sent to you in error and the e-mail contains patient
information, please contact the Partners Compliance HelpLine at
http://www.partners.org/complianceline . If the e-mail was sent to you in error
but does not contain patient information, please contact the sender and
properly dispose of the e-mail.
_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer_______________________________________________
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer The information in
this e-mail is intended only for the person to whom it is addressed. If you
believe this e-mail was sent to you in error and the e-mail contains patient
information, please contact the Partners Compliance HelpLine at
http://www.partners.org/complianceline . If the e-mail was sent to you in error
but does not contain patient information, please contact the sender and
properly dispose of the e-mail.
_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer_______________________________________________
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer 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
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.
