2
$\begingroup$

I am studying the impact of temperature on the onset of flowering of trees. I am using a cox model for time dependent covariates as implemented in the R package survival. I only have one (cumulative) time dependent covariate, that is daily growing degree days (GDD, calculated from daily min and max temperature).

My results look something like this:

     coef exp(coef)  se(coef)     z Pr(>|z|)      
0.0117237 1.0117927 0.0004184 28.02   <2e-16 ***

Now my question: I understand a hazard ratio of 1.01 would mean that a unit increase in my time dependent covariate increases the risk of death by 1%, however, in my case all trees are expected to flower.

What I want to know is whether a unit increase in my time dependent covariates causes earlier flowering. I this information contained in the hazard ratio, or am I doing something wrong?

$\endgroup$
2
$\begingroup$

Whether or not all observations are uncensored does not affect the interpretation of a hazard ratio. The hazard ratio is an instantaneous event rate, roughly speaking the probability of having an event within a small interval of time around time $t$ divided by the probability of not having the event before time $t$. Hazard ratios are ratios of such instantaneous hazard rates. So a baseline or time-dependent covariate can modify the hazard rate of the ultimate outcome.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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