Take the 2-minute tour ×
Cross Validated is a question and answer site for people interested in statistics, machine learning, data analysis, data mining, and data visualization. It's 100% free, no registration required.

I'm working on a survival analysis project. In part of the project, I'm asked to "identify factors that may contribute to an increased risk of death". I have difficult time understanding what I'm asked for. Could anyone help me understand this? How should I identify those factors? Is this an example of building exploratory model? If so, should I only pick variables that increase the risk?

Thanks for your help.

share|improve this question
    
Is this a homework assignment? –  gung Dec 11 '12 at 2:18
    
Hi @gung, no, it's actually a project of my survival analysis course. I don't need any hint since it's not fair to other students, I just need to understand what I'm asked in this question. –  user48405 Dec 11 '12 at 2:37
    
Can you give us a better idea of what you have to work with? For example, are you just being asked to design a project, or are you working from a data set, etc.? –  Fomite Dec 11 '12 at 2:50
    
hi @EpiGrad, I have a dataset and the project has many parts. I have developed my model which is the relationship between BMI and risk of death for post-MI patients. Now the last question is the one I mentioned above. –  user48405 Dec 11 '12 at 2:53
add comment

2 Answers

up vote 3 down vote accepted

Doesn't this just mean to put your available covariates into your model and test for significant effects? Of course it depends on whether you have an accelerated failure time or proportional hazard model, as to the direction of the significant effects, but that's how I read it.

share|improve this answer
1  
The question reads an increased risk of death - the effect estimates need to not only be significant, but significant and positive. –  Fomite Dec 11 '12 at 2:56
add comment

I think @CoreySparks is essentially right here. The key is that either your instructor wants you to conduct one-tailed tests, or just report back the variables that are both 'significant' (via a two-tailed test), and tend to increase death rates.

share|improve this answer
1  
Yes, and the direction of the effect will depend on the type of model, for an AFT (think weibull model) a positive coefficient indicates a longer survival time, or a decrease in risk of death, versus, say a Cox PH model where a positive values means an increase in the hazard of death, which is what I interpret the OP as asking about. –  Corey Sparks Dec 11 '12 at 15:59
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.