Re: Odds Ratios - univariate to multivariate

Hi,

All these resources are great, I've now idetified my problem, I don't
have one variable causing the change but rather 3. One which is very
important.
What I know is that this variable is not as strongly related to my DV
as the other variables, but is very strongly related to these IVs.
Which makes sense, because I don't think that residents in a nursing
home who cry more are more likely to wander, however they appear to
have other problem behaviours, such as insomnia, which can increase
wandering behaviour.
Now I have been reading about this, but it is mostly looking at
multiple linear regression. What I want to know is, should I leave this
variable in and explain it as a negative suppressor variable, or should
I remove it? With it in the model the ORs are stronger, but not by a
huge amount.
There are also other variables I am wishing to add, should I see if any
of these effect the model with and without this variable in the model?
Thank you so much for your help!

Bellinda

.