[R] discrepancies between stata and r for a cox regression

Achim Zeileis Achim.Zeileis at wu-wien.ac.at
Mon May 18 18:19:44 CEST 2009

On Mon, 18 May 2009, Michel Boutsen wrote:

> Hello
> I would like to develop the use of R.
> Trying R and more particulary the cox model, I am surprised by discrepancies between results with stata and R for a cox model
> With the same data base, I get a hazard ratio (4.82) that is not the same obtained with stata (4.52)

I would expect that setting method = "breslow" replicates the results from 
Stata (et al.). As ?coxph points out:

   method: a character string specifying the method for tie handling.
           If there   are no tied death times all the methods are
           equivalent.  Nearly all Cox regression programs use the
           Breslow method by default,  but not this one.  The Efron
           approximation is used as the default here, as it is much more
            accurate when dealing with tied death times, and is as
           efficient  computationally.  The exact method computes the
           exact partial likelihood, which is  equivalent  to a
           conditional logistic model.  If there are a large number of
           ties  the computational time will be excessive.


> You will find attached the file leukemia.dta I used (Stata)
> Here are the codes for R
> leukemia=read.fwf(file="leukem.txt",widths=c(4,2,3,2,5,2),col.names=c("id","TREAT","TIME","STATUS","LOGWBC", "GENDER"))
> library(survival)
> res <- coxph(Surv(TIME, STATUS)~TREAT, data=leukemia)
> summary(res)
> and here the codes for for stata
> infix ID 2-3 TREAT 6 TIME 8-9 STATUS 11 LOGWBC 12-16 GENDER 18 using "g:rleukem.txt",clear
> stset TIME, failure(STATUS==1)
> stcox TREAT
> SPSS and EPIinfo give the same HR than Stata
> I tried with an other database without any problem
> What would be the problem??? I changed of pc and versions of R (2.81 & 2.9.0) without any change. The means are the same for the two packages.
> I saw a few posts with discrepancies but not with the same database
> Thanks in advance
> Michel Boutsen
> Brussel's University
> Department of Biostatistics

More information about the R-help mailing list