spssforum.com

Hi
I am struggling with Cox Regression Analysis (PASW Statistics 18) & would appreciate your help. I am a PhD student and am evaluating the association between malnutrition in >3000 hospital patients & health-related outcomes such as mortality and hospital length of stay.
The question I want to answer is "Does the hazard of in-hospital mortality differ between well-nourished and malnourished patients?" I have an extensive dataset and from the literature I have identified confounding factors (eg type of disease, severity of disease, age, etc). However, there are other factors that I want to control for (not previously evaluated in the literature; example participants' food intake). When I enter all the confounding factors (those from the literature and those that I want to evaluate), the output is strange (i.e. expB is between 10,000-35000 for some variables, some variables have empty hazard ratio cells).

How do I identify which variables should be entered into the model?? Are there any diagnostics that I can run in spss to decide this??
I would much appreciate your help
Thanks