Re: [Freesurfer] Eliminate variable which is the same for all patients from FreeSurfer's Design Matrix
Hi Kai, please remember to post to the list. Your FSGD file is not quite right. Gender is a discrete variable and should be represented by two groups not as a covariate. If Sequence is discrete, then you need four groups (Gender by Sequence).
doug
On 3/9/18 3:20 AM, k.ohmstedt@stud.uni-heidelberg.de wrote:
Dear Douglas,
the fsgd file I used is attached. Thank you for your help!
Best regards
Kai
Zitat von "Douglas N. Greve" dgreve@mgh.harvard.edu:
can you send your fsgd file so that I have a better idea of what you are mentioning?
On 03/08/2018 08:39 AM, k.ohmstedt@stud.uni-heidelberg.de wrote:
Dear all,
I am trying to correlate psychometric measurements and the local Gyrification Index. To do so, I use the FreeSurfer pipeline to calculate the lGI and then PALM, following the advice in this thread (https://mail.nmr.mgh.harvard.edu/pipermail//freesurfer/2017-March/050703.htm...).
All my subjects are part of the same group, so I used a FSGD with the group "main" to create the design matrix and mask for my data that are required by PALM. Having a closer look at the design matrix that was created, I found that there was a variable for the group that was the same for all my patients. As it is the same for all patients, I thought eliminating it would not be a problem. But after re-running PALM without it, there were huge differences in my results and effects were notably larger and more significant.
Do any of you have any experience which option is best in this case? Is it a valid choice to eliminate the variable for the group, as it is the same for each patient anyway?
Furthermore, would you recommend centering for the design matrix? I found that this can have an impact, but I am lost on in which cases it should be done and in which it shouldn't.
Thank you for your help!
Best regards
Kai Ohmstedt
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.
Hi Douglas,
thanks for the advice.
I have one more question: I want to find correlations of the lGI and the respective psychometric measurement, regressing out the effect of gender, sequence and the other covariates. I just had a look at the examples of FSGDs containing two factors with two levels each and as far as I understand it, it is only possible to compare two of the groups or to find an interaction between groups and covariates.
Is it at all possible to find out how the lGI is correlated to one of my covariates, i.e. higher values of Beck's Depression Inventory correlate with higher/lower values of the lGI, when regressing out the other factors? If yes, how do I build the contrast file to do so correctly? I am stuck here.
Thank you for the time and effort!
Best regards
Kai
Zitat von Douglas Greve dgreve@mgh.harvard.edu:
Hi Kai, please remember to post to the list. Your FSGD file is not quite right. Gender is a discrete variable and should be represented by two groups not as a covariate. If Sequence is discrete, then you need four groups (Gender by Sequence).
doug
On 3/9/18 3:20 AM, k.ohmstedt@stud.uni-heidelberg.de wrote:
Dear Douglas,
the fsgd file I used is attached. Thank you for your help!
Best regards
Kai
Zitat von "Douglas N. Greve" dgreve@mgh.harvard.edu:
can you send your fsgd file so that I have a better idea of what you are mentioning?
On 03/08/2018 08:39 AM, k.ohmstedt@stud.uni-heidelberg.de wrote:
Dear all,
I am trying to correlate psychometric measurements and the local Gyrification Index. To do so, I use the FreeSurfer pipeline to calculate the lGI and then PALM, following the advice in this thread (https://mail.nmr.mgh.harvard.edu/pipermail//freesurfer/2017-March/050703.htm...). All my subjects are part of the same group, so I used a FSGD with the group "main" to create the design matrix and mask for my data that are required by PALM. Having a closer look at the design matrix that was created, I found that there was a variable for the group that was the same for all my patients. As it is the same for all patients, I thought eliminating it would not be a problem. But after re-running PALM without it, there were huge differences in my results and effects were notably larger and more significant.
Do any of you have any experience which option is best in this case? Is it a valid choice to eliminate the variable for the group, as it is the same for each patient anyway?
Furthermore, would you recommend centering for the design matrix? I found that this can have an impact, but I am lost on in which cases it should be done and in which it shouldn't.
Thank you for your help!
Best regards
Kai Ohmstedt
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.
Yes, let's say you have 4 groups (eg, gender and sequence) and one covariate (BDI score). If you use a DODS model, then you will have 8 regressors (4 intercepts and 4 slopes). To test for an effect of covariate regressing out the group, then you would have
[0 0 0 0 .25 .25 .25 .25]
The +0.25 computes the mean slope across all groups (note you could also just use all 1s, you will get the same p-value).
On 03/09/2018 01:43 PM, k.ohmstedt@stud.uni-heidelberg.de wrote:
Hi Douglas,
thanks for the advice.
I have one more question: I want to find correlations of the lGI and the respective psychometric measurement, regressing out the effect of gender, sequence and the other covariates. I just had a look at the examples of FSGDs containing two factors with two levels each and as far as I understand it, it is only possible to compare two of the groups or to find an interaction between groups and covariates.
Is it at all possible to find out how the lGI is correlated to one of my covariates, i.e. higher values of Beck's Depression Inventory correlate with higher/lower values of the lGI, when regressing out the other factors? If yes, how do I build the contrast file to do so correctly? I am stuck here.
Thank you for the time and effort!
Best regards
Kai
Zitat von Douglas Greve dgreve@mgh.harvard.edu:
Hi Kai, please remember to post to the list. Your FSGD file is not quite right. Gender is a discrete variable and should be represented by two groups not as a covariate. If Sequence is discrete, then you need four groups (Gender by Sequence).
doug
On 3/9/18 3:20 AM, k.ohmstedt@stud.uni-heidelberg.de wrote:
Dear Douglas,
the fsgd file I used is attached. Thank you for your help!
Best regards
Kai
Zitat von "Douglas N. Greve" dgreve@mgh.harvard.edu:
can you send your fsgd file so that I have a better idea of what you are mentioning?
On 03/08/2018 08:39 AM, k.ohmstedt@stud.uni-heidelberg.de wrote:
Dear all,
I am trying to correlate psychometric measurements and the local Gyrification Index. To do so, I use the FreeSurfer pipeline to calculate the lGI and then PALM, following the advice in this thread (https://mail.nmr.mgh.harvard.edu/pipermail//freesurfer/2017-March/050703.htm...). All my subjects are part of the same group, so I used a FSGD with the group "main" to create the design matrix and mask for my data that are required by PALM. Having a closer look at the design matrix that was created, I found that there was a variable for the group that was the same for all my patients. As it is the same for all patients, I thought eliminating it would not be a problem. But after re-running PALM without it, there were huge differences in my results and effects were notably larger and more significant.
Do any of you have any experience which option is best in this case? Is it a valid choice to eliminate the variable for the group, as it is the same for each patient anyway?
Furthermore, would you recommend centering for the design matrix? I found that this can have an impact, but I am lost on in which cases it should be done and in which it shouldn't.
Thank you for your help!
Best regards
Kai Ohmstedt
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.
Hi Douglas,
thank you for the explanation of the contrasts, this helped a lot!
All the best,
Kai
2018-03-12 18:20 GMT+01:00 Douglas N. Greve dgreve@mgh.harvard.edu:
Yes, let's say you have 4 groups (eg, gender and sequence) and one covariate (BDI score). If you use a DODS model, then you will have 8 regressors (4 intercepts and 4 slopes). To test for an effect of covariate regressing out the group, then you would have
[0 0 0 0 .25 .25 .25 .25]
The +0.25 computes the mean slope across all groups (note you could also just use all 1s, you will get the same p-value).
On 03/09/2018 01:43 PM, k.ohmstedt@stud.uni-heidelberg.de wrote:
Hi Douglas,
thanks for the advice.
I have one more question: I want to find correlations of the lGI and the respective psychometric measurement, regressing out the effect of gender, sequence and the other covariates. I just had a look at the examples of FSGDs containing two factors with two levels each and as far as I understand it, it is only possible to compare two of the groups or to find an interaction between groups and covariates.
Is it at all possible to find out how the lGI is correlated to one of my covariates, i.e. higher values of Beck's Depression Inventory correlate with higher/lower values of the lGI, when regressing out the other factors? If yes, how do I build the contrast file to do so correctly? I am stuck here.
Thank you for the time and effort!
Best regards
Kai
Zitat von Douglas Greve dgreve@mgh.harvard.edu:
Hi Kai, please remember to post to the list. Your FSGD file is not
quite right. Gender is a discrete variable and should be represented by two groups not as a covariate. If Sequence is discrete, then you need four groups (Gender by Sequence).
doug
On 3/9/18 3:20 AM, k.ohmstedt@stud.uni-heidelberg.de wrote:
Dear Douglas,
the fsgd file I used is attached. Thank you for your help!
Best regards
Kai
Zitat von "Douglas N. Greve" dgreve@mgh.harvard.edu:
can you send your fsgd file so that I have a better idea of what you are
mentioning?
On 03/08/2018 08:39 AM, k.ohmstedt@stud.uni-heidelberg.de wrote:
Dear all,
I am trying to correlate psychometric measurements and the local Gyrification Index. To do so, I use the FreeSurfer pipeline to calculate the lGI and then PALM, following the advice in this thread (https://mail.nmr.mgh.harvard.edu/pipermail//freesurfer/2017 -March/050703.html). All my subjects are part of the same group, so I used a FSGD with the group "main" to create the design matrix and mask for my data that are required by PALM. Having a closer look at the design matrix that was created, I found that there was a variable for the group that was the same for all my patients. As it is the same for all patients, I thought eliminating it would not be a problem. But after re-running PALM without it, there were huge differences in my results and effects were notably larger and more significant.
Do any of you have any experience which option is best in this case? Is it a valid choice to eliminate the variable for the group, as it is the same for each patient anyway?
Furthermore, would you recommend centering for the design matrix? I found that this can have an impact, but I am lost on in which cases it should be done and in which it shouldn't.
Thank you for your help!
Best regards
Kai Ohmstedt
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.
Stop sending me this shit
Wiadomość napisana przez Kai Ohmstedt kai.ohmstedt@googlemail.com w dniu 13.03.2018, o godz. 02:02:
Hi Douglas,
thank you for the explanation of the contrasts, this helped a lot!
All the best,
Kai
2018-03-12 18:20 GMT+01:00 Douglas N. Greve dgreve@mgh.harvard.edu:
Yes, let's say you have 4 groups (eg, gender and sequence) and one covariate (BDI score). If you use a DODS model, then you will have 8 regressors (4 intercepts and 4 slopes). To test for an effect of covariate regressing out the group, then you would have
[0 0 0 0 .25 .25 .25 .25]
The +0.25 computes the mean slope across all groups (note you could also just use all 1s, you will get the same p-value).
On 03/09/2018 01:43 PM, k.ohmstedt@stud.uni-heidelberg.de wrote: Hi Douglas,
thanks for the advice.
I have one more question: I want to find correlations of the lGI and the respective psychometric measurement, regressing out the effect of gender, sequence and the other covariates. I just had a look at the examples of FSGDs containing two factors with two levels each and as far as I understand it, it is only possible to compare two of the groups or to find an interaction between groups and covariates.
Is it at all possible to find out how the lGI is correlated to one of my covariates, i.e. higher values of Beck's Depression Inventory correlate with higher/lower values of the lGI, when regressing out the other factors? If yes, how do I build the contrast file to do so correctly? I am stuck here.
Thank you for the time and effort!
Best regards
Kai
Zitat von Douglas Greve dgreve@mgh.harvard.edu:
Hi Kai, please remember to post to the list. Your FSGD file is not quite right. Gender is a discrete variable and should be represented by two groups not as a covariate. If Sequence is discrete, then you need four groups (Gender by Sequence).
doug
On 3/9/18 3:20 AM, k.ohmstedt@stud.uni-heidelberg.de wrote: Dear Douglas,
the fsgd file I used is attached. Thank you for your help!
Best regards
Kai
Zitat von "Douglas N. Greve" dgreve@mgh.harvard.edu:
can you send your fsgd file so that I have a better idea of what you are mentioning?
> On 03/08/2018 08:39 AM, k.ohmstedt@stud.uni-heidelberg.de wrote: > Dear all, > > I am trying to correlate psychometric measurements and the local > Gyrification Index. > To do so, I use the FreeSurfer pipeline to calculate the lGI and then > PALM, following the advice in this thread > (https://mail.nmr.mgh.harvard.edu/pipermail//freesurfer/2017-March/050703.htm...). All my subjects are part of the same group, so I used a FSGD with the > group "main" to create the design matrix and mask for my data that are > required by PALM. > Having a closer look at the design matrix that was created, I found > that there was a variable for the group that was the same for all my > patients. As it is the same for all patients, I thought eliminating it > would not be a problem. But after re-running PALM without it, there > were huge differences in my results and effects were notably larger > and more significant. > > Do any of you have any experience which option is best in this case? > Is it a valid choice to eliminate the variable for the group, as it is > the same for each patient anyway? > > Furthermore, would you recommend centering for the design matrix? I > found that this can have an impact, but I am lost on in which cases it > should be done and in which it shouldn't. > > > Thank you for your help! > > > Best regards > > Kai Ohmstedt > > > _______________________________________________ > 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@nmr.mgh.harvard.edu
-
Douglas Greve -
Douglas N. Greve
-
k.ohmstedt＠stud.uni-heidelberg.de -
Kai Ohmstedt -
Weronika Jęchorek