Michael,
I'm also not aware of any program that will automatically demean by group when you request an interaction or when you add a covariate.
I still think its important to stress that if you demean the entire group, you are looking at the "covariate-adjusted group means" rather than the actual group means. There are clearly times when you might want to compare group means and control for variance associated with a variable. One example that comes to mind is AD versus normal controls and MMSE scores. For this hypothetical, lets say the mean MMSE across the entire group was 27. You wouldn't want to compare AD subjects with a mean of 27 to controls with a mean of 27 as AD subjects very rarely have an MMSE score of 27 as well as controls rarely having and MMSE score of 27. Nevertheless, you might want to account for the variance associated with MMSE scores within each group.
Neither solution is right or wrong, it just needs to be stated that the comparison is of covariate-adjusted means/intercepts/means in papers.
Best Regards, Donald McLaren ================= D.G. McLaren, Ph.D. Research Fellow, Department of Neurology, Massachusetts General Hospital and Harvard Medical School Postdoctoral Research Fellow, GRECC, Bedford VA Website: http://www.martinos.org/~mclaren Office: (773) 406-2464 ===================== This e-mail contains CONFIDENTIAL INFORMATION which may contain PROTECTED HEALTHCARE INFORMATION and may also be LEGALLY PRIVILEGED and which is intended only for the use of the individual or entity named above. If the reader of the e-mail is not the intended recipient or the employee or agent responsible for delivering it to the intended recipient, you are hereby notified that you are in possession of confidential and privileged information. Any unauthorized use, disclosure, copying or the taking of any action in reliance on the contents of this information is strictly prohibited and may be unlawful. If you have received this e-mail unintentionally, please immediately notify the sender via telephone at (773) 406-2464 or email.
On Mon, Nov 26, 2012 at 5:53 PM, Michael Harms mharms@conte.wustl.edu wrote:
Just wanted to mention that, to my knowledge no commercial stat package will mean center by group (Donald's case (C)) if you request an interaction model. Mean centering by group is a very unusual operation.
cheers, -MH
-- Michael Harms, Ph.D.
Conte Center for the Neuroscience of Mental Disorders Washington University School of Medicine Department of Psychiatry, Box 8134 660 South Euclid Ave. Tel: 314-747-6173 St. Louis, MO 63110 Email: mharms@wustl.edu
On 11/26/12 4:32 PM, "MCLAREN, Donald" mclaren.donald@gmail.com wrote:
I assume that your 12 columns are: your 6 groups followed by 6 columns representing your covariate.
You must first test that the slopes are not different with an F-test: 0 0 0 0 0 0 1 -1 0 0 0 0 0 0 0 0 0 0 0 1 -1 0 0 0 0 0 0 0 0 0 0 0 1 -1 0 0 0 0 0 0 0 0 0 0 0 1 -1 0
If you don't find any differences, then you can use the DOSS model. If you find differences, then you have an interaction between your covariate and groups. When this happens, you should consult a statistician if you want to interpret the effects of group as the general rule of thumb is not to interpret the main effects in the presence of an interaction. Using the DOSS model will leave you with 7 regressors. The test for group differences will be: 1 -1 0 0 0 0 0 or 0 1 -1 0 0 0 0 etc. depending on what you are wanting to test. Simply having the covariate in the model corrects for age.
A couple of points on covariates - for more details see (http://mumford.fmripower.org/mean_centering/): (A) If you don't mean center, then you will be looking at the group intercepts; (B) If you mean center, then you will be looking at the covariate-adjusted group means; (C) If you mean center by group, then you will be looking at the group means and reducing the variance due to age. For example, if you want to compare old and young subjects, but control for age within each group.
A and B will give you the same answer in the DOSS since the slopes of each group are parallel. C will give you a different result. Depending on how you want to interpret your results will effect which method you want to use for mean-centering or not-mean centering the data. In all cases, mean-centering will not effect the slope estimates/significance.
Hope this clears up the confusion.
Best Regards, Donald McLaren
D.G. McLaren, Ph.D. Research Fellow, Department of Neurology, Massachusetts General Hospital and Harvard Medical School Postdoctoral Research Fellow, GRECC, Bedford VA Website: http://www.martinos.org/~mclaren Office: (773) 406-2464 ===================== This e-mail contains CONFIDENTIAL INFORMATION which may contain PROTECTED HEALTHCARE INFORMATION and may also be LEGALLY PRIVILEGED and which is intended only for the use of the individual or entity named above. If the reader of the e-mail is not the intended recipient or the employee or agent responsible for delivering it to the intended recipient, you are hereby notified that you are in possession of confidential and privileged information. Any unauthorized use, disclosure, copying or the taking of any action in reliance on the contents of this information is strictly prohibited and may be unlawful. If you have received this e-mail unintentionally, please immediately notify the sender via telephone at (773) 406-2464 or email.
On Fri, Nov 23, 2012 at 2:01 AM, 李仕广 imaging_shiguangli@163.com wrote:
Dear FreeSurfers: I hope you can help me out with a problem in disgning my .mtx file. I want to examine the difference among three groups (HC group, EX group and PT group). Considing gender, I have six groups as similar as the example described in http://surfer.nmr.mgh.harvard.edu/fswiki/Fsgdf6G0V. But I want to correct for age. Then, I have 6 classes (two factors/three l! evels) a nd one covariate (age). How can I compare the difference among three groups (HC group, EX group, and PT group)? Could you please tall me how to design the contrast matrix with 12 regressors?
Thanks a zillion times!
Best wishes
shiguangli
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