[R] ROC curve with survival data

[Frank E Harrell Jr]

SUBIRANA CACHINERO, ISAAC wrote:
> Hi everyone, 
> 
> I am doing 5 years mortality predictive index score with survival analysis using a Cox proportional hazard model where I have a continous predictive variable and a right censored response which is the mortality, and the individuals were followed a maximum of 7 years.
> 
> I'd like to asses the discrimination ability of survival analysis Cox model by computing a ROC curve and area under the curve for a fixed time (5 years), taking into account that the response is not binary but right censored.
> 
> So, is there a function that computes a ROC curve under right censored survival data?
> 
> Thank you in advance.

I don't find ROC curves themselves very useful but the area under them 
is useful and is a simple translation of the Somers' Dxy rank 
correlation between predicted survival probability (or anything 
monotonically related to it just is log hazard) and observed survival 
time.  Dxy = 2*(C-.5) where C is the concordance index, a generalization 
of ROC area not requiring choosing a specific time point.  You can get 
this in the rcorr.cens function in the Hmisc package.

Frank

> 
>  
> 
>  
> 
> Isaac Subirana: isubirana at imim.es
> 
> Institut Municipal d'Investigació Mèdica (IMIM),
> 
> Barcelona (Spain)
> 
>  
> 
> 
> 	[[alternative HTML version deleted]]
> 
> 
> 
> ------------------------------------------------------------------------
> 
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html


-- 
Frank E Harrell Jr   Professor and Chair           School of Medicine
                      Department of Biostatistics   Vanderbilt University