correlated variables

#1
Hello,

In a Cox model, if we have several variables correlated with each other, is this a problem in the results? Should we add interactions or not?
What is the solution with variables correlated?

Thanks
 
Last edited:

Karabiner

TS Contributor
#2
is this a problem in the results?
Since your description could mean e.g. r=0.1 as well as r=0.98, and since we do not know the model,
or the variables, or the research question, or the sample size, your question cannot be answered
specifically.
Generally speaking, it could sometimes cause interpretation problems with regard to the contribution
of the respective variables, especially if one variable can be mostly explained by one or several other
variables (multicollinearity).
Should we add interactions or not?
Without knowing the model, or the variables, or the research question, or the sample size, I find it
a bit too difficult to comment on this.
What is the solution with variables correlated?
I am not sure what you mean. Could you elaborate this a bit more?

With kind regards

Karabiner