Dear Freesurfer experts,
I would like to study the relationship between cortical thickness and one clinical variable with qdec, but correcting for age and gender.
Given that I have only one group and 2 covariates (one continuous, one dichotomic) I don't know how I should set the design of my analysis in qdec, especially for the gender variable.
I have tried two different ways (both DOSS design):
1) selecting Discrete = gender; Continuous = clinical measure; Nuisance factor = age and looking at the contrast called : "Does the correlation between thickness and clinical measure accounting for gender differ from 0? nuisance factor : age"
2) selecting : continuous = clinical measure; Nuisance Factor = age, gender (coded as 1 and 2) and looking at the contrast called : "Does the correlation between thickness and clinical measure differ from 0, nuisance factor : age, gender"
But the two contrasts give very different results, which I find very surprising. I exported cortical thickness at the peak significance of the clusters and tried to run a GLM myself in SPSS and it seems that coding gender as a continuous variable with two values (1 and 2) provides the most realistic results. However, I saw many times on the mailing list that you recommend to use gender as a discrete variable, so I am very confused. Could you explain me the difference between these contrasts and help me to identify which one will accurately identify the correlation between cortical thickness and my clinical variable correcting for the effect of age and gender.
Thank you in advance for your answer, Best regards Elisa
Hi Elisa, don't use the DOSS feature in QDEC. Sorry, I sent out an email about 6mo ago on this, but it is not easy to let people know about a bug once the bug is out there. doug
On 09/19/2013 11:30 AM, E. Scariati wrote:
Dear Freesurfer experts,
I would like to study the relationship between cortical thickness and one clinical variable with qdec, but correcting for age and gender.
Given that I have only one group and 2 covariates (one continuous, one dichotomic) I don't know how I should set the design of my analysis in qdec, especially for the gender variable.
I have tried two different ways (both DOSS design):
- selecting Discrete = gender; Continuous = clinical measure;
Nuisance factor = age and looking at the contrast called : "Does the correlation between thickness and clinical measure accounting for gender differ from 0? nuisance factor : age"
- selecting : continuous = clinical measure; Nuisance Factor = age,
gender (coded as 1 and 2) and looking at the contrast called : "Does the correlation between thickness and clinical measure differ from 0, nuisance factor : age, gender"
But the two contrasts give very different results, which I find very surprising. I exported cortical thickness at the peak significance of the clusters and tried to run a GLM myself in SPSS and it seems that coding gender as a continuous variable with two values (1 and 2) provides the most realistic results. However, I saw many times on the mailing list that you recommend to use gender as a discrete variable, so I am very confused. Could you explain me the difference between these contrasts and help me to identify which one will accurately identify the correlation between cortical thickness and my clinical variable correcting for the effect of age and gender.
Thank you in advance for your answer, Best regards Elisa
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Hi Doug,
I'm jumping in the discussion because I was a bit scared with your previous email mentioning that this DOSS bug affects all FreeSurfer's versions. Does that also affect statistical analyses computed with mri_glmfit using the command line? Do you have an insight whether the bias introduced by the bug is important or not? (as others may also be, I'm becoming a bit anxious about previously published results…)
Finally, to get back to Elisa's question: do you have some suggestion in the mean time to assess the relationship between cortical thickness and a clinical measure correcting for age and gender? Using DODS? With or without demeaning the covariates and nuisance?
Sorry for the abundance of questions, and, as always, thanks a lot for your answer!
Marie
On Sep 20, 2013, at 6:13 PM, Douglas N Greve greve@nmr.mgh.harvard.edu wrote:
Hi Elisa, don't use the DOSS feature in QDEC. Sorry, I sent out an email about 6mo ago on this, but it is not easy to let people know about a bug once the bug is out there. doug
On 09/19/2013 11:30 AM, E. Scariati wrote:
Dear Freesurfer experts,
I would like to study the relationship between cortical thickness and one clinical variable with qdec, but correcting for age and gender.
Given that I have only one group and 2 covariates (one continuous, one dichotomic) I don't know how I should set the design of my analysis in qdec, especially for the gender variable.
I have tried two different ways (both DOSS design):
- selecting Discrete = gender; Continuous = clinical measure;
Nuisance factor = age and looking at the contrast called : "Does the correlation between thickness and clinical measure accounting for gender differ from 0? nuisance factor : age"
- selecting : continuous = clinical measure; Nuisance Factor = age,
gender (coded as 1 and 2) and looking at the contrast called : "Does the correlation between thickness and clinical measure differ from 0, nuisance factor : age, gender"
But the two contrasts give very different results, which I find very surprising. I exported cortical thickness at the peak significance of the clusters and tried to run a GLM myself in SPSS and it seems that coding gender as a continuous variable with two values (1 and 2) provides the most realistic results. However, I saw many times on the mailing list that you recommend to use gender as a discrete variable, so I am very confused. Could you explain me the difference between these contrasts and help me to identify which one will accurately identify the correlation between cortical thickness and my clinical variable correcting for the effect of age and gender.
Thank you in advance for your answer, Best regards Elisa
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 greve@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.
Hi Marie,
On 9/20/13 4:30 PM, Marie Schaer wrote:
Hi Doug,
I'm jumping in the discussion because I was a bit scared with your previous email mentioning that this DOSS bug affects all FreeSurfer's versions. Does that also affect statistical analyses computed with mri_glmfit using the command line? Do you have an insight whether the bias introduced by the bug is important or not? (as others may also be, I'm becoming a bit anxious about previously published results…)
It does not affect the command-line version, only in QDEC. It was basically not creating a contrast matrix that matched the hypothesis question under some circumstances.
Finally, to get back to Elisa's question: do you have some suggestion in the mean time to assess the relationship between cortical thickness and a clinical measure correcting for age and gender? Using
DODS? With or without demeaning the covariates and nuisance? Sorry for the abundance of questions, and, as always, thanks a lot for your answer! Marie
I would probably do it with DODS and just test the mean across the two groups, eg, Class M Class F Variables ClinicalVar Age
[0 0 .5 .5 0 0]
This would account for possible differences in slope between M and F. In the end, I think it will give you about the same as if use DOSS. If you have a small sample size, you could use DOSS because DODS will cost you 2 more DOF
doug
On Sep 20, 2013, at 6:13 PM, Douglas N Greve greve@nmr.mgh.harvard.edu wrote:
Hi Elisa, don't use the DOSS feature in QDEC. Sorry, I sent out an email about 6mo ago on this, but it is not easy to let people know about a bug once the bug is out there. doug
On 09/19/2013 11:30 AM, E. Scariati wrote:
Dear Freesurfer experts,
I would like to study the relationship between cortical thickness and one clinical variable with qdec, but correcting for age and gender.
Given that I have only one group and 2 covariates (one continuous, one dichotomic) I don't know how I should set the design of my analysis in qdec, especially for the gender variable.
I have tried two different ways (both DOSS design):
- selecting Discrete = gender; Continuous = clinical measure;
Nuisance factor = age and looking at the contrast called : "Does the correlation between thickness and clinical measure accounting for gender differ from 0? nuisance factor : age"
- selecting : continuous = clinical measure; Nuisance Factor = age,
gender (coded as 1 and 2) and looking at the contrast called : "Does the correlation between thickness and clinical measure differ from 0, nuisance factor : age, gender"
But the two contrasts give very different results, which I find very surprising. I exported cortical thickness at the peak significance of the clusters and tried to run a GLM myself in SPSS and it seems that coding gender as a continuous variable with two values (1 and 2) provides the most realistic results. However, I saw many times on the mailing list that you recommend to use gender as a discrete variable, so I am very confused. Could you explain me the difference between these contrasts and help me to identify which one will accurately identify the correlation between cortical thickness and my clinical variable correcting for the effect of age and gender.
Thank you in advance for your answer, Best regards Elisa
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 greve@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.
Hello Doug, Hello Marie,
thank you very much for your answers, I'm sorry I had missed your previous mail about QDEC!
I will use mri_glmfit then to confirm my results,
Best, Elisa
2013/9/20 Douglas Greve greve@nmr.mgh.harvard.edu
Hi Marie,
On 9/20/13 4:30 PM, Marie Schaer wrote:
Hi Doug,
I'm jumping in the discussion because I was a bit scared with your
previous email mentioning that this DOSS bug affects all FreeSurfer's versions. Does that also affect statistical analyses computed with mri_glmfit using the command line? Do you have an insight whether the bias introduced by the bug is important or not? (as others may also be, I'm becoming a bit anxious about previously published results…) It does not affect the command-line version, only in QDEC. It was basically not creating a contrast matrix that matched the hypothesis question under some circumstances.
Finally, to get back to Elisa's question: do you have some suggestion in
the mean time to assess the relationship between cortical thickness and a clinical measure correcting for age and gender? Using DODS? With or without demeaning the covariates and nuisance? Sorry for the abundance of questions, and, as always, thanks a lot for your answer! Marie
I would probably do it with DODS and just test the mean across the two groups, eg, Class M Class F Variables ClinicalVar Age
[0 0 .5 .5 0 0]
This would account for possible differences in slope between M and F. In the end, I think it will give you about the same as if use DOSS. If you have a small sample size, you could use DOSS because DODS will cost you 2 more DOF
doug
On Sep 20, 2013, at 6:13 PM, Douglas N Greve greve@nmr.mgh.harvard.edu wrote:
Hi Elisa, don't use the DOSS feature in QDEC. Sorry, I sent out an email about 6mo ago on this, but it is not easy to let people know about a bug once the bug is out there. doug
On 09/19/2013 11:30 AM, E. Scariati wrote:
Dear Freesurfer experts,
I would like to study the relationship between cortical thickness and one clinical variable with qdec, but correcting for age and gender.
Given that I have only one group and 2 covariates (one continuous, one dichotomic) I don't know how I should set the design of my analysis in qdec, especially for the gender variable.
I have tried two different ways (both DOSS design):
- selecting Discrete = gender; Continuous = clinical measure;
Nuisance factor = age and looking at the contrast called : "Does the correlation between thickness and clinical measure accounting for gender differ from 0? nuisance factor : age"
- selecting : continuous = clinical measure; Nuisance Factor = age,
gender (coded as 1 and 2) and looking at the contrast called : "Does the correlation between thickness and clinical measure differ from 0, nuisance factor : age, gender"
But the two contrasts give very different results, which I find very surprising. I exported cortical thickness at the peak significance of the clusters and tried to run a GLM myself in SPSS and it seems that coding gender as a continuous variable with two values (1 and 2) provides the most realistic results. However, I saw many times on the mailing list that you recommend to use gender as a discrete variable, so I am very confused. Could you explain me the difference between these contrasts and help me to identify which one will accurately identify the correlation between cortical thickness and my clinical variable correcting for the effect of age and gender.
Thank you in advance for your answer, Best regards Elisa
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 greve@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
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@nmr.mgh.harvard.edu
-
Douglas Greve -
Douglas N Greve
-
E. Scariati -
Marie Schaer