[R] time dependent hazard ratios
David Winsemius
dwinsemius at comcast.net
Mon Apr 18 22:49:14 CEST 2011
On Apr 18, 2011, at 4:09 PM, array chip wrote:
> Sorry this is a re-post. I posted it last night, haven't heard from
> anyone, hope
> this moves the thread up a little and anyone can comment?
My observation is that coefficients of 19 in Cox (or other exponential
models) models generally indicate numerical difficulties rather than
correct estimates.
--
David.
>
> Thanks!
>
> John
>
>
>
> ----- Forwarded Message ----
> From: array chip <arrayprofile at yahoo.com>
> To: r-help at r-project.org
> Sent: Sun, April 17, 2011 11:33:32 PM
> Subject: time dependent hazard ratios
>
>
> Hi, I am new to time-dependent Cox model to estimate time dependent
> hazard
> ratios. Let me use aml dataset from survival package as an example:
>
>> aml3<-survSplit(aml,cut=c(5,10,20),end="time",start="start",
> event="status",episode="i")
>
> This will generate the new data frame based on the time cut points
> 5, 10 and 20
> and produce a indicator variable "i" that indicates the time
> interval membership
> for each observation.
>
> If I want to esimate hazard ratio of variable "x" (Nonmaintained vs
> Maintained)
> for each of the time intervals 0-5, 5-10, 10-20 and >=20, would the
> following
> calculate such HRs?
>
>> coxph(Surv(start,time,status)~x*as.factor(i),data=aml3)
> Call:
> coxph(formula = Surv(start, time, status) ~ x * as.factor(i),
> data = aml3)
>
>
> coef exp(coef) se(coef) z p
> xNonmaintained 19.2 2.10e+08 10461 0.00183 1
> as.factor(i)1 NA NA 0 NA NA
> as.factor(i)2 NA NA 0 NA NA
> as.factor(i)3 NA NA 0 NA NA
> xNonmaintained:as.factor(i)1 -18.3 1.17e-08 10461 -0.00175 1
> xNonmaintained:as.factor(i)2 -19.6 3.18e-09 10461 -0.00187 1
> xNonmaintained:as.factor(i)3 -18.0 1.46e-08 10461 -0.00172 1
>
> So my understanding is that:
> for time interval 0-5, HR=exp(19.2)=2.1e+08
> for time interval 5-10, HR=exp(19.2-18.3)=2.46
> for time interval 10-20, HR=exp(19.2-19.6)=0.67
> for time interval >=20, HR=exp(19.2-18.0)=3.32
>
> Am I correct on estimating these time dependent HRs?
>
> Thanks
>
> John
>
> [[alternative HTML version deleted]]
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
David Winsemius, MD
West Hartford, CT
More information about the R-help
mailing list