You should model sex as a categorical factor, not as a continuous variable, so there is no issue of centering sex. Whether you really need to center age depends on what you want to test. If you want to test of a difference in age slopes regressing out the effects of sex, then you don't need to change the age values. If you want to test for a difference in sex regressing out the effects of age, then you should first test for a difference in age slopes as described above. If there are no differences, then you should re-run the analysis using DOSS (different offset, same slope) in which case you do not need to demean the age. doug
On 10/05/2012 04:58 PM, Yang, Daniel wrote:
Dear freesurfer experts,
I'm using Qdec and I want to control for age (a continuous variable) and sex (f,m or 1,0).
Do I need to center age (i.e., subtract the score from its mean) before I select it as a nuisance factor? Do I need to center sex?
I found that selecting age_centered as a nuisance factor vs. selecting age_uncentered as a nuisance factor led to very different results… not sure what is the correct way to go. In brief, if I want to control for age and sex, which of the following is the correct way to go?
- Center neither age nor sex (f,m). Then choose age_uncentered as a nuisance factor, but sex_uncentered as a discrete/fixed factor.
- Center age but not sex (f,m). Then choose age_centered as a nuisance factor, but sex_uncentered as a discrete/fixed factor.
- Center age but not sex (1,0). Then choose both age_centered and sex_uncentered as nuisance factors.
- Center age and sex (1,0). Then choose both age_centered and sex_centered as nuisance factors.
Thanks in advance! Daniel -- Yung-Jui "Daniel" Yang, PhD Postdoctoral Researcher Yale Child Study Center New Haven, CT (203) 737-5454
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer