nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

871542_en-US

liliconcom

Multicolliniarity issue in a Cox Proportional Hazard Model

Hi Everybody,

I need help to understand and address the following issue.

I'm building a model to predict the development of cardiovascular disease in a cohort. I’ve included two variables that are correlated with each other (R² = 0.83). When I include both in the model, one of them flips direction—it becomes protective instead of a risk factor. But when I include each variable separately, both show a positive association with disease, as expected.

I don't think this is due to a biological explanation—it seems more like a multicollinearity issue. I know I could use either VIF or PCA to address this. But since my outcome is binary, I’m unsure how to correctly calculate VIF. I tried running a least squares regression between the two variables and got a VIF of 1, but I don’t think that’s the right approach here.

Has anyone dealt with a similar situation? How would you recommend addressing this?

Advanced Statistical Modeling

Predictive Modeling and Machine Learning

Windows

Re: Multicolliniarity issue in a Cox Proportional Hazard Model

Hi @RafaelZS ,

Have you considered simply placing an interaction term (X1*X2) in the model to see the effect on the model outcomes and the VIF values? If you could provide a dataset that would be helpful.