Dear Experts,
I have 2 groups - controls (22 subjects) and patients (19 subjects) with mean ages that are not significantly different, though ranges/median are. I want to compare cortical thickness between groups while allowing for age (age as nuisance factor) but also want to see if there is an interaction between age and group (age as continuous variable). I have demeaned the age across both controls and patients and then used qdec.
1) Using DODS and demeaned ages I find that there is an age-gp interaction such that patients have quicker thickness decline with age than controls. This survives FDR 0.05 and monte carlo VWP 0.005 CWP 0.05. Is this finding valid given that I used demeaned ages (rather than absolute ages) to look for interaction or does it make no difference ?
2) When I come to look to see whether there is a gp difference in thickness I enter age as a nuisance variable in qdec (as opposed to continuous variable above) and use DODS with demeaned age variable across gp. Am I safe to interpret my findings normally given age-gp interaction or is is not safe - if it is not safe how can I look for gp differences or it ok to interpret results provided gp differences are not in regions where I see an interaction ?
Thanks.
Mahinda
On 10/16/2012 02:08 PM, Mahinda Yogarajah wrote:
Dear Experts,
I have 2 groups - controls (22 subjects) and patients (19 subjects) with mean ages that are not significantly different, though ranges/median are. I want to compare cortical thickness between groups while allowing for age (age as nuisance factor) but also want to see if there is an interaction between age and group (age as continuous variable). I have demeaned the age across both controls and patients and then used qdec.
- Using DODS and demeaned ages I find that there is an age-gp interaction such that patients have quicker thickness decline with age than controls. This survives FDR 0.05 and monte carlo VWP 0.005 CWP 0.05. Is this finding valid given that I used demeaned ages (rather than absolute ages) to look for interaction or does it make no difference ?
For this contrast (group-slope interaction), it makes no difference.
- When I come to look to see whether there is a gp difference in thickness I enter age as a nuisance variable in qdec (as opposed to continuous variable above) and use DODS with demeaned age variable across gp. Am I safe to interpret my findings normally given age-gp interaction or is is not safe - if it is not safe how can I look for gp differences or it ok to interpret results provided gp differences are not in regions where I see an interaction ?
It is not safe. If there is an interaction, then the difference between groups changes with age. You can think of it as two thickness vs age lines, one for each group. If there is no interaction, then these lines will be parallel, meaning that the difference between the groups will be the same no matter what age. If there is an interaction, these lines will not be parallel, meaning that the group difference changes with age, and so whether you see a difference between them or not depends on what age you test them at. Without demeaning, this test age will be age=0. If you demean, then the age will be age=mean(age).
doug
Thanks.
Mahinda
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
freesurfer@nmr.mgh.harvard.edu
-
Douglas N Greve -
Mahinda Yogarajah