In this case you would have six regressors: 1. control-offset 2. hiv-offset 3. control-age 4. hiv-age 5. control-ed 6. hiv-ed
If you want to test group thickness "regressing out" the effect of age and ed, then you would just set the contrast to:
1 -1 0 0 0 0 0
Note that this will test it at age=0 and ed=0, which you may or may not want to do. Ideally, you would test them at some age/ed in the range of your sample. A lot of people simply demean each covariate, but this can bias the results depending upon who you add or remove. Better to choose an age/ed level independent of your sample.
doug
On Sat, 13 Jun 2009, Jeff Sadino wrote:
Hello Freesurfers,
My apologies for this very basic question more about statistics than freesurfer, but I have not found an answer anywhere on the wikis, or the web (maybe it is too basic :-) ) I am interested in age-related effects of hiv and I wanted to covary for age and education in mri_glmfit. For example, if my fsgd file is:
class control class hiv variables age ed input subj1 control 40 13 input subj2 hiv 35 12 ... ...
and I use dods, is my contrast matrix the place where I would tell it to covary? Something like: 0 0 -1 1 1 1 1 1 (y-int) (age slope) (covary for age) (covary for ed)
As always, thank you for your help, Jeff
Hotmail® has ever-growing storage! Don?t worry about storage limits. http://windowslive.com/Tutorial/Hotmail/Storage?ocid=TXT_TAGLM_WL_HM_Tutoria...