Dear Freesurfer's experts,
I have searched through the mailing list but haven't found any answer to my question.
Is it possible with LME model to make correlations between for example cortical thickness surface data and cognition scores along time ? As it is possible to test for interaction of group X time, is this also in the same way feasible to test for coognition score X time on cortical thickness ?
Many thanks for your advice !
Best regards, Matthieu
Hi Matthieu,
yes, that is possible. Instead of group, you use a variable for your score (and interaction etc). Sometimes it may also makes sense to use score instead of time.
Best, Martin
On 12 May 2017, at 10:51, Matthieu Vanhoutte matthieuvanhoutte@gmail.com wrote:
Dear Freesurfer's experts,
I have searched through the mailing list but haven't found any answer to my question.
Is it possible with LME model to make correlations between for example cortical thickness surface data and cognition scores along time ? As it is possible to test for interaction of group X time, is this also in the same way feasible to test for coognition score X time on cortical thickness ?
Many thanks for your advice !
Best regards, Matthieu _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Hi Martin,
Thank you. How should this variable be coded ? Should it be as age covariate where age at baseline is used along all time points of each subject ?
Could you provide me an example of design matrix, I don't manage to see what does it look like to.
Best regards, Matthieu
Le 14 mai 2017 8:08 PM, "Martin Reuter" mreuter@nmr.mgh.harvard.edu a écrit :
Hi Matthieu,
yes, that is possible. Instead of group, you use a variable for your score (and interaction etc). Sometimes it may also makes sense to use score instead of time.
Best, Martin
On 12 May 2017, at 10:51, Matthieu Vanhoutte matthieuvanhoutte@gmail.com
wrote:
Dear Freesurfer's experts,
I have searched through the mailing list but haven't found any answer to
my question.
Is it possible with LME model to make correlations between for example
cortical thickness surface data and cognition scores along time ? As it is possible to test for interaction of group X time, is this also in the same way feasible to test for coognition score X time on cortical thickness ?
Many thanks for your advice !
Best regards, Matthieu _______________________________________________ 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 Matthieu,
one option is to replace time with score in the model. That should be straight forward.
The other option is to include score as a time-varying covariate. If your score is not varying much across time and you are more interested if the average score (or baseline score) affects atrophy rates, you can also include is as a standard (fixed across time) co-variate (such as baseline age) with a time (and a group etc) interaction.
Sorry, but I cannot do your design. Ultimately the model is the research question that you are asking and it is important that this is done correctly. Maybe there is a local biostats person that you can talk to?
Best, Martin
On 05/15/2017 08:20 PM, Matthieu Vanhoutte wrote:
Hi Martin,
Thank you. How should this variable be coded ? Should it be as age covariate where age at baseline is used along all time points of each subject ?
Could you provide me an example of design matrix, I don't manage to see what does it look like to.
Best regards, Matthieu
Le 14 mai 2017 8:08 PM, "Martin Reuter" <mreuter@nmr.mgh.harvard.edu mailto:mreuter@nmr.mgh.harvard.edu> a écrit :
Hi Matthieu, yes, that is possible. Instead of group, you use a variable for your score (and interaction etc). Sometimes it may also makes sense to use score instead of time. Best, Martin > On 12 May 2017, at 10:51, Matthieu Vanhoutte <matthieuvanhoutte@gmail.com <mailto:matthieuvanhoutte@gmail.com>> wrote: > > Dear Freesurfer's experts, > > I have searched through the mailing list but haven't found any answer to my question. > > Is it possible with LME model to make correlations between for example cortical thickness surface data and cognition scores along time ? As it is possible to test for interaction of group X time, is this also in the same way feasible to test for coognition score X time on cortical thickness ? > > Many thanks for your advice ! > > Best regards, > Matthieu > _______________________________________________ > Freesurfer mailing list > Freesurfer@nmr.mgh.harvard.edu <mailto:Freesurfer@nmr.mgh.harvard.edu> > https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer <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 <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 <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
Hi Martin,
Thank you for this detailed answer.
Are replacing time by score or include score as time-varying covariate leading to the same result because of looking at the same effect ?
My willing would be to find patterns of atrophy rate/progression correlated with cognitive score. In context of AD which method would be the best for you ?
Best, Matthieu
Le 16 mai 2017 11:17 AM, "Martin Reuter" mreuter@nmr.mgh.harvard.edu a écrit :
Hi Matthieu,
one option is to replace time with score in the model. That should be straight forward.
The other option is to include score as a time-varying covariate. If your score is not varying much across time and you are more interested if the average score (or baseline score) affects atrophy rates, you can also include is as a standard (fixed across time) co-variate (such as baseline age) with a time (and a group etc) interaction.
Sorry, but I cannot do your design. Ultimately the model is the research question that you are asking and it is important that this is done correctly. Maybe there is a local biostats person that you can talk to?
Best, Martin
On 05/15/2017 08:20 PM, Matthieu Vanhoutte wrote:
Hi Martin,
Thank you. How should this variable be coded ? Should it be as age covariate where age at baseline is used along all time points of each subject ?
Could you provide me an example of design matrix, I don't manage to see what does it look like to.
Best regards, Matthieu
Le 14 mai 2017 8:08 PM, "Martin Reuter" mreuter@nmr.mgh.harvard.edu a écrit :
Hi Matthieu,
yes, that is possible. Instead of group, you use a variable for your score (and interaction etc). Sometimes it may also makes sense to use score instead of time.
Best, Martin
On 12 May 2017, at 10:51, Matthieu Vanhoutte matthieuvanhoutte@gmail.com
wrote:
Dear Freesurfer's experts,
I have searched through the mailing list but haven't found any answer to
my question.
Is it possible with LME model to make correlations between for example
cortical thickness surface data and cognition scores along time ? As it is possible to test for interaction of group X time, is this also in the same way feasible to test for coognition score X time on cortical thickness ?
Many thanks for your advice !
Best regards, Matthieu _______________________________________________ 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 listFreesurfer@nmr.mgh.harvard.eduhttps://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 Matthieu,
replacing time by score is very different from adding score as a covariate. Often scores are crude and often they are constant in controls (always full score), and only vary slightly in diseased. In those cases it may not be good to use score as a time variable.
I would either add avg. score as a variable (fixed across time - would probably do this for intial testing anyway) or score as time-varying covariate, but would read up on how to interpret results in the presence of time-varying-covariates. I am not a statistician.
Best, Martin
On 05/16/2017 08:09 PM, Matthieu Vanhoutte wrote:
Hi Martin,
Thank you for this detailed answer.
Are replacing time by score or include score as time-varying covariate leading to the same result because of looking at the same effect ?
My willing would be to find patterns of atrophy rate/progression correlated with cognitive score. In context of AD which method would be the best for you ?
Best, Matthieu
Le 16 mai 2017 11:17 AM, "Martin Reuter" <mreuter@nmr.mgh.harvard.edu mailto:mreuter@nmr.mgh.harvard.edu> a écrit :
Hi Matthieu, one option is to replace time with score in the model. That should be straight forward. The other option is to include score as a time-varying covariate. If your score is not varying much across time and you are more interested if the average score (or baseline score) affects atrophy rates, you can also include is as a standard (fixed across time) co-variate (such as baseline age) with a time (and a group etc) interaction. Sorry, but I cannot do your design. Ultimately the model is the research question that you are asking and it is important that this is done correctly. Maybe there is a local biostats person that you can talk to? Best, Martin On 05/15/2017 08:20 PM, Matthieu Vanhoutte wrote:Hi Martin, Thank you. How should this variable be coded ? Should it be as age covariate where age at baseline is used along all time points of each subject ? Could you provide me an example of design matrix, I don't manage to see what does it look like to. Best regards, Matthieu Le 14 mai 2017 8:08 PM, "Martin Reuter" <mreuter@nmr.mgh.harvard.edu <mailto:mreuter@nmr.mgh.harvard.edu>> a écrit : Hi Matthieu, yes, that is possible. Instead of group, you use a variable for your score (and interaction etc). Sometimes it may also makes sense to use score instead of time. Best, Martin > On 12 May 2017, at 10:51, Matthieu Vanhoutte <matthieuvanhoutte@gmail.com <mailto:matthieuvanhoutte@gmail.com>> wrote: > > Dear Freesurfer's experts, > > I have searched through the mailing list but haven't found any answer to my question. > > Is it possible with LME model to make correlations between for example cortical thickness surface data and cognition scores along time ? As it is possible to test for interaction of group X time, is this also in the same way feasible to test for coognition score X time on cortical thickness ? > > Many thanks for your advice ! > > Best regards, > Matthieu > _______________________________________________ > Freesurfer mailing list > Freesurfer@nmr.mgh.harvard.edu <mailto:Freesurfer@nmr.mgh.harvard.edu> > https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer <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 <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 <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 <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 <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 <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
Hi Martin,
I will read up on interpretation of time varying covariates.
If initially I use score as variable fixed across time, and define a variable for 'score x time' interaction: 1) Would putting only 1 to 'score x time' column (for contrast) test for progression of correlation patterns between atrophy and score ? 2) How to interpret this according to sign of contrast (knowing that decrease in score means worse performance) ?
Many thanks for your time and lights !
Best, Matthieu
Le 17 mai 2017 1:46 PM, "Martin Reuter" mreuter@nmr.mgh.harvard.edu a écrit :
Hi Matthieu,
replacing time by score is very different from adding score as a covariate. Often scores are crude and often they are constant in controls (always full score), and only vary slightly in diseased. In those cases it may not be good to use score as a time variable.
I would either add avg. score as a variable (fixed across time - would probably do this for intial testing anyway) or score as time-varying covariate, but would read up on how to interpret results in the presence of time-varying-covariates. I am not a statistician. Best, Martin
On 05/16/2017 08:09 PM, Matthieu Vanhoutte wrote:
Hi Martin,
Thank you for this detailed answer.
Are replacing time by score or include score as time-varying covariate leading to the same result because of looking at the same effect ?
My willing would be to find patterns of atrophy rate/progression correlated with cognitive score. In context of AD which method would be the best for you ?
Best, Matthieu
Le 16 mai 2017 11:17 AM, "Martin Reuter" mreuter@nmr.mgh.harvard.edu a écrit :
Hi Matthieu,
one option is to replace time with score in the model. That should be straight forward.
The other option is to include score as a time-varying covariate. If your score is not varying much across time and you are more interested if the average score (or baseline score) affects atrophy rates, you can also include is as a standard (fixed across time) co-variate (such as baseline age) with a time (and a group etc) interaction.
Sorry, but I cannot do your design. Ultimately the model is the research question that you are asking and it is important that this is done correctly. Maybe there is a local biostats person that you can talk to?
Best, Martin
On 05/15/2017 08:20 PM, Matthieu Vanhoutte wrote:
Hi Martin,
Thank you. How should this variable be coded ? Should it be as age covariate where age at baseline is used along all time points of each subject ?
Could you provide me an example of design matrix, I don't manage to see what does it look like to.
Best regards, Matthieu
Le 14 mai 2017 8:08 PM, "Martin Reuter" mreuter@nmr.mgh.harvard.edu a écrit :
Hi Matthieu,
yes, that is possible. Instead of group, you use a variable for your score (and interaction etc). Sometimes it may also makes sense to use score instead of time.
Best, Martin
On 12 May 2017, at 10:51, Matthieu Vanhoutte matthieuvanhoutte@gmail.com
wrote:
Dear Freesurfer's experts,
I have searched through the mailing list but haven't found any answer to
my question.
Is it possible with LME model to make correlations between for example
cortical thickness surface data and cognition scores along time ? As it is possible to test for interaction of group X time, is this also in the same way feasible to test for coognition score X time on cortical thickness ?
Many thanks for your advice !
Best regards, Matthieu _______________________________________________ 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 listFreesurfer@nmr.mgh.harvard.eduhttps://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
_______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.e du/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/compli anceline . 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 listFreesurfer@nmr.mgh.harvard.eduhttps://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 Matthieu,
1) you would put a column of score and a column of score X time. The first allows you to test if the intercept changes based on scores (e.g. if hippocampal volume is affected by the score, controlling for whatever else you included, e.g. age and gender etc) and the second interaction allows you to test if the the slopes are affected by score (so if atrophy rates are different at different score levels).
2) if your effect for that interaction column is positive, there is a positive correlation (higher score, slower atrophy (or even brain growth), and lower score with faster atrophy - this is what you would expect) if there is a negative effect, it would mean the opposite.
Best, Martin
On 05/17/2017 11:12 PM, Matthieu Vanhoutte wrote:
Hi Martin,
I will read up on interpretation of time varying covariates.
If initially I use score as variable fixed across time, and define a variable for 'score x time' interaction:
- Would putting only 1 to 'score x time' column (for contrast) test
for progression of correlation patterns between atrophy and score ? 2) How to interpret this according to sign of contrast (knowing that decrease in score means worse performance) ?
Many thanks for your time and lights !
Best, Matthieu
Le 17 mai 2017 1:46 PM, "Martin Reuter" <mreuter@nmr.mgh.harvard.edu mailto:mreuter@nmr.mgh.harvard.edu> a écrit :
Hi Matthieu, replacing time by score is very different from adding score as a covariate. Often scores are crude and often they are constant in controls (always full score), and only vary slightly in diseased. In those cases it may not be good to use score as a time variable. I would either add avg. score as a variable (fixed across time - would probably do this for intial testing anyway) or score as time-varying covariate, but would read up on how to interpret results in the presence of time-varying-covariates. I am not a statistician. Best, Martin On 05/16/2017 08:09 PM, Matthieu Vanhoutte wrote:Hi Martin, Thank you for this detailed answer. Are replacing time by score or include score as time-varying covariate leading to the same result because of looking at the same effect ? My willing would be to find patterns of atrophy rate/progression correlated with cognitive score. In context of AD which method would be the best for you ? Best, Matthieu Le 16 mai 2017 11:17 AM, "Martin Reuter" <mreuter@nmr.mgh.harvard.edu <mailto:mreuter@nmr.mgh.harvard.edu>> a écrit : Hi Matthieu, one option is to replace time with score in the model. That should be straight forward. The other option is to include score as a time-varying covariate. If your score is not varying much across time and you are more interested if the average score (or baseline score) affects atrophy rates, you can also include is as a standard (fixed across time) co-variate (such as baseline age) with a time (and a group etc) interaction. Sorry, but I cannot do your design. Ultimately the model is the research question that you are asking and it is important that this is done correctly. Maybe there is a local biostats person that you can talk to? Best, Martin On 05/15/2017 08:20 PM, Matthieu Vanhoutte wrote:Hi Martin, Thank you. How should this variable be coded ? Should it be as age covariate where age at baseline is used along all time points of each subject ? Could you provide me an example of design matrix, I don't manage to see what does it look like to. Best regards, Matthieu Le 14 mai 2017 8:08 PM, "Martin Reuter" <mreuter@nmr.mgh.harvard.edu <mailto:mreuter@nmr.mgh.harvard.edu>> a écrit : Hi Matthieu, yes, that is possible. Instead of group, you use a variable for your score (and interaction etc). Sometimes it may also makes sense to use score instead of time. Best, Martin > On 12 May 2017, at 10:51, Matthieu Vanhoutte <matthieuvanhoutte@gmail.com <mailto:matthieuvanhoutte@gmail.com>> wrote: > > Dear Freesurfer's experts, > > I have searched through the mailing list but haven't found any answer to my question. > > Is it possible with LME model to make correlations between for example cortical thickness surface data and cognition scores along time ? As it is possible to test for interaction of group X time, is this also in the same way feasible to test for coognition score X time on cortical thickness ? > > Many thanks for your advice ! > > Best regards, > Matthieu > _______________________________________________ > Freesurfer mailing list > Freesurfer@nmr.mgh.harvard.edu <mailto:Freesurfer@nmr.mgh.harvard.edu> > https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer <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 <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 <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 <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 <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 <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 <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 <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 <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
Hi Martin,
Thank you this helps !
Please find last question below inline.
Best, Matthieu
Le 18 mai 2017 10:33 AM, "Martin Reuter" mreuter@nmr.mgh.harvard.edu a écrit :
Hi Matthieu,
1) you would put a column of score and a column of score X time. The first allows you to test if the intercept changes based on scores (e.g. if hippocampal volume is affected by the score, controlling for whatever else you included, e.g. age and gender etc) and the second interaction allows you to test if the the slopes are affected by score (so if atrophy rates are different at different score levels).
2) if your effect for that interaction column is positive, there is a positive correlation (higher score, slower atrophy (or even brain growth), and lower score with faster atrophy - this is what you would expect) if there is a negative effect, it would mean the opposite.
How do you deduce this from combination of atrophy, score and time ? I know that positive leads to positive correlation and when group X time is negative means that cortical thickness decreases with time. However I can't manage to understand relation between cortical thickness and score X time. Could you explain me ?
Best, Martin
On 05/17/2017 11:12 PM, Matthieu Vanhoutte wrote:
Hi Martin,
I will read up on interpretation of time varying covariates.
If initially I use score as variable fixed across time, and define a variable for 'score x time' interaction: 1) Would putting only 1 to 'score x time' column (for contrast) test for progression of correlation patterns between atrophy and score ? 2) How to interpret this according to sign of contrast (knowing that decrease in score means worse performance) ?
Many thanks for your time and lights !
Best, Matthieu
Le 17 mai 2017 1:46 PM, "Martin Reuter" mreuter@nmr.mgh.harvard.edu a écrit :
Hi Matthieu,
replacing time by score is very different from adding score as a covariate. Often scores are crude and often they are constant in controls (always full score), and only vary slightly in diseased. In those cases it may not be good to use score as a time variable.
I would either add avg. score as a variable (fixed across time - would probably do this for intial testing anyway) or score as time-varying covariate, but would read up on how to interpret results in the presence of time-varying-covariates. I am not a statistician. Best, Martin
On 05/16/2017 08:09 PM, Matthieu Vanhoutte wrote:
Hi Martin,
Thank you for this detailed answer.
Are replacing time by score or include score as time-varying covariate leading to the same result because of looking at the same effect ?
My willing would be to find patterns of atrophy rate/progression correlated with cognitive score. In context of AD which method would be the best for you ?
Best, Matthieu
Le 16 mai 2017 11:17 AM, "Martin Reuter" mreuter@nmr.mgh.harvard.edu a écrit :
Hi Matthieu,
one option is to replace time with score in the model. That should be straight forward.
The other option is to include score as a time-varying covariate. If your score is not varying much across time and you are more interested if the average score (or baseline score) affects atrophy rates, you can also include is as a standard (fixed across time) co-variate (such as baseline age) with a time (and a group etc) interaction.
Sorry, but I cannot do your design. Ultimately the model is the research question that you are asking and it is important that this is done correctly. Maybe there is a local biostats person that you can talk to?
Best, Martin
On 05/15/2017 08:20 PM, Matthieu Vanhoutte wrote:
Hi Martin,
Thank you. How should this variable be coded ? Should it be as age covariate where age at baseline is used along all time points of each subject ?
Could you provide me an example of design matrix, I don't manage to see what does it look like to.
Best regards, Matthieu
Le 14 mai 2017 8:08 PM, "Martin Reuter" mreuter@nmr.mgh.harvard.edu a écrit :
Hi Matthieu,
yes, that is possible. Instead of group, you use a variable for your score (and interaction etc). Sometimes it may also makes sense to use score instead of time.
Best, Martin
On 12 May 2017, at 10:51, Matthieu Vanhoutte matthieuvanhoutte@gmail.com
wrote:
Dear Freesurfer's experts,
I have searched through the mailing list but haven't found any answer to
my question.
Is it possible with LME model to make correlations between for example
cortical thickness surface data and cognition scores along time ? As it is possible to test for interaction of group X time, is this also in the same way feasible to test for coognition score X time on cortical thickness ?
Many thanks for your advice !
Best regards, Matthieu _______________________________________________ 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 listFreesurfer@nmr.mgh.harvard.eduhttps://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
_______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.e du/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/compli anceline . 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 listFreesurfer@nmr.mgh.harvard.eduhttps://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
_______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.e du/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/compli anceline . 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 listFreesurfer@nmr.mgh.harvard.eduhttps://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 Matthieu,
Y = b0 + b1 tij + b2 si + b3 si * tij
(Y thickness, b0 intercept, b1 slope, b2 score, b3 score effect on slope, where t_ij is the time of subject i at time point j, s_i is subject score)
= (b0 + b2 si) + (b1 + b3 si) tij
so the slope is made up of (b1 + b3 si), a global component and a subject component that adds or subtracts to the slope based on the score. If the b3 is positive, the slope gets larger with increasing score. Here larger means closer to zero if the slope was negative, so less atrophy or even brain growth.
If b3 is negative, the slope gets smaller with increasing score: more atrophy or less brain growth, depending on b1.
Best Martin
On 05/18/2017 01:12 PM, Matthieu Vanhoutte wrote:
Hi Martin,
Thank you this helps !
Please find last question below inline.
Best, Matthieu
Le 18 mai 2017 10:33 AM, "Martin Reuter" <mreuter@nmr.mgh.harvard.edu mailto:mreuter@nmr.mgh.harvard.edu> a écrit :
Hi Matthieu, 1) you would put a column of score and a column of score X time. The first allows you to test if the intercept changes based on scores (e.g. if hippocampal volume is affected by the score, controlling for whatever else you included, e.g. age and gender etc) and the second interaction allows you to test if the the slopes are affected by score (so if atrophy rates are different at different score levels). 2) if your effect for that interaction column is positive, there is a positive correlation (higher score, slower atrophy (or even brain growth), and lower score with faster atrophy - this is what you would expect) if there is a negative effect, it would mean the opposite.How do you deduce this from combination of atrophy, score and time ? I know that positive leads to positive correlation and when group X time is negative means that cortical thickness decreases with time. However I can't manage to understand relation between cortical thickness and score X time. Could you explain me ?
Best, Martin On 05/17/2017 11:12 PM, Matthieu Vanhoutte wrote:Hi Martin, I will read up on interpretation of time varying covariates. If initially I use score as variable fixed across time, and define a variable for 'score x time' interaction: 1) Would putting only 1 to 'score x time' column (for contrast) test for progression of correlation patterns between atrophy and score ? 2) How to interpret this according to sign of contrast (knowing that decrease in score means worse performance) ? Many thanks for your time and lights ! Best, Matthieu Le 17 mai 2017 1:46 PM, "Martin Reuter" <mreuter@nmr.mgh.harvard.edu <mailto:mreuter@nmr.mgh.harvard.edu>> a écrit : Hi Matthieu, replacing time by score is very different from adding score as a covariate. Often scores are crude and often they are constant in controls (always full score), and only vary slightly in diseased. In those cases it may not be good to use score as a time variable. I would either add avg. score as a variable (fixed across time - would probably do this for intial testing anyway) or score as time-varying covariate, but would read up on how to interpret results in the presence of time-varying-covariates. I am not a statistician. Best, Martin On 05/16/2017 08:09 PM, Matthieu Vanhoutte wrote:Hi Martin, Thank you for this detailed answer. Are replacing time by score or include score as time-varying covariate leading to the same result because of looking at the same effect ? My willing would be to find patterns of atrophy rate/progression correlated with cognitive score. In context of AD which method would be the best for you ? Best, Matthieu Le 16 mai 2017 11:17 AM, "Martin Reuter" <mreuter@nmr.mgh.harvard.edu <mailto:mreuter@nmr.mgh.harvard.edu>> a écrit : Hi Matthieu, one option is to replace time with score in the model. That should be straight forward. The other option is to include score as a time-varying covariate. If your score is not varying much across time and you are more interested if the average score (or baseline score) affects atrophy rates, you can also include is as a standard (fixed across time) co-variate (such as baseline age) with a time (and a group etc) interaction. Sorry, but I cannot do your design. Ultimately the model is the research question that you are asking and it is important that this is done correctly. Maybe there is a local biostats person that you can talk to? Best, Martin On 05/15/2017 08:20 PM, Matthieu Vanhoutte wrote:Hi Martin, Thank you. How should this variable be coded ? Should it be as age covariate where age at baseline is used along all time points of each subject ? Could you provide me an example of design matrix, I don't manage to see what does it look like to. Best regards, Matthieu Le 14 mai 2017 8:08 PM, "Martin Reuter" <mreuter@nmr.mgh.harvard.edu <mailto:mreuter@nmr.mgh.harvard.edu>> a écrit : Hi Matthieu, yes, that is possible. Instead of group, you use a variable for your score (and interaction etc). Sometimes it may also makes sense to use score instead of time. Best, Martin > On 12 May 2017, at 10:51, Matthieu Vanhoutte <matthieuvanhoutte@gmail.com <mailto:matthieuvanhoutte@gmail.com>> wrote: > > Dear Freesurfer's experts, > > I have searched through the mailing list but haven't found any answer to my question. > > Is it possible with LME model to make correlations between for example cortical thickness surface data and cognition scores along time ? As it is possible to test for interaction of group X time, is this also in the same way feasible to test for coognition score X time on cortical thickness ? > > Many thanks for your advice ! > > Best regards, > Matthieu > _______________________________________________ > Freesurfer mailing list > Freesurfer@nmr.mgh.harvard.edu <mailto:Freesurfer@nmr.mgh.harvard.edu> > https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer <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 <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 <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 <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 <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 <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 <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 <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 <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 <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 <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 <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@nmr.mgh.harvard.edu
-
Martin Reuter -
Matthieu Vanhoutte