[R] missing values and survival analysis

[Frank E Harrell Jr]

On Sun, 11 Apr 2004 23:37:22 -0400
john.ferguson at yale.edu wrote:

> 
> Hi everyone,
> 
> I'm analysing a survival analysis data set at the moment with missing 
> values in the covariate and survival vectors (I have about 60 
> variables).  I know there are some functions on the CRAN network to 
> deal with missing values in general multivariate data.  Does anybody 
> know of any package that deals with missing data specifically in the 
> context of survival analysis.  Any help would be greatly appreciated.
> 
> Thanks,
> 
> john.

Consider using the aregImpute function in the Hmisc package with
right-censored survival times by predicting baseline covariates using the
follow-up time, event indicator/censoring, and the product of the two,
using multiple imputation.  I am not comfortable with imputing follow-up
time and event indicators from covariates though.  If the follow-up time
is completely missing you might consider discarding the observation.

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