ANOVA question

[edieb]

---

Members don't see this ad.

I have a question about ANOVAs. I was reading a journal article where they had 3 independent variables: Autism + Depression, PDD-NOS + Depression and Depression only.

I thought that including depression in all 3 independent variables would control for the variance contributed by depression. However, someone told me that including depression in all 3 ind variable DOES NOT control for the variance because depression still impacts the dependent variable.

Is this true?

 

[psychanon]

Well, your question really depends on what the analysis was trying to show-- are they trying to control for depression as a nuisance variable or are they showing how different diagnostic categories differ on whatever outcome. If you are asking whether you can disregard depression because all three categories had depression, the answer is likely no, because depression, though often assessed categorically, is not really a categorical construct. It is possible that depression levels differ systematically between groups, and if fact that is likely because depression levels tend to be higher when depression is comorbid with another disorder compared to non-comorbid depression, although I have no idea whether that applies to PDD. So, if groups 1 and 2 have more depression than group 3, and depression affects the DV, then no, depression is not effectively controlled for here.

 

[myelin]

Can you not just use ANCOVA with depression as the covariate? Wouldn't this control for the effects that depression has on the other variables?

 

[edieb]

Both of you have good points. They talked about the possibility of using ANCOVA but stated that since they were trying to control for the variance contributed by depression and see the predictive values of the other variables, they used sequential regression and looked at the Beta weights entering dpx as block 1 and then the other two disorders

 

[JockNerd]

Look at procedures to eliminate multicollinearity (e.g. regressing out). I'm pretty sure that's what you're trying to do, from reading your post.

Edit: Duh, maybe I should read more carefully. That's what *they* should have done.