Regression Covariables

[#1]

Quick question. I am using a Cox regression model and was wondering if I could include variables that are dependent upon each other in the same model. For example, would adding height, weight, and BMI as covariables affect the model, given that BMI is directly dependent upon height and weight? Another example, in transplant studies I want to know how gender affects survival. So I use both recipient and donor gender as covariables. But I also want to look at gender match/mismatch (M:M, M:F, F:M, and F:F). Can I incorporate all three variables, recipient gender, donor gender, and gender match/mismatch, in the same Cox regression model? Thanks

[#2]

Basically does interacting covariables have any impact on the Cox regression model as a whole?

[#3]

I would not include both BMI and the weight and height in the same model. Linear models should avoid collinearity if possible. You could evaluate two models, one using BMI and another using weight and height instead of BMI.

From your brief description of the variables, it sounds reasonable to include mismatch as well. That information should be independent of gender, no?

[#4]

In addition, if you have JMP Pro, Generalized Regression platform support Cox Proportional Hazards with advanced variable selection techniques.

Generalized Regression Documentation