[R-sig-Epi] cox model with time-dependent covariates

[BXC (Bendix Carstensen)]

You do not necessarily have to have one observation per time-window if you use a Cox mode, you just need a new record every time the time-dependent variable changes.

You would presumably get a better feel for you data and an easier access to the analyses if you used the Lexis facilities in the Epi package.
Two papers have recently appeared in Journal of Statistical Software 
http://www.jstatsoft.org/, describing this:

Lexis: An R Class for Epidemiological Studies with Long-Term Follow-Up
Vol. 38, Issue 5, Jan 2011
Authors: Martyn Plummer, Bendix Carstensen

Using Lexis Objects for Multi-State Models in R
Vol. 38, Issue 6, Jan 2011
Authors: Bendix Carstensen, Martyn Plummer
 
Best regrads,
Bendix Carstensen

> -----Original Message-----
> From: r-sig-epi-bounces at r-project.org 
> [mailto:r-sig-epi-bounces at r-project.org] On Behalf Of raphaëlle métras
> Sent: 17. januar 2011 13:23
> To: r-sig-epi at r-project.org
> Subject: [R-sig-Epi] cox model with time-dependent covariates
> 
> Dear all,
> 
> I am planning to run a cox model with time-dependent 
> covariates. I would like to use the 'fold' function (Fox, 2002:
> http://cran.r-project.org/doc/contrib/Fox-Companion/appendix-c
ox-regression.pdf)
> to create the dataset, with one observation per subject & 
> time-window, and then  use the 'coxph' function of the 
> survival package to run the model.
> 
> However, I had a couple of questions, and I can't find the 
> answer on the
> internet:
> 
> 1. In the document writtent by Fox, the new dataset created 
> for the purpose of the time-dependent covariates model 
> (Rossi.2) numbers the observations 1.1, 1.2, 1.3, 1.4, 1.5, 
> etc. which corresponds to the measurements in the different 
> time-windows for subject 1. My question is the following: if 
> I run a cox model on such a dataset, is the model accounting 
> for the fact that 1.1, 1.2, 1.3, ... belongs to the same subject?
> 
> 2. I will have a big dataset, i.e. 40,000 subjects and 36 
> time-windows, i.e. 1,440,000 lines, and about 20 covariates. 
> I was wondering if I should expect some big troubles to run 
> my analysis, or can the software handle so many rows?
> 
> Many thanks for your help and your time,
> 
> Regards,
> 
> Raphaelle
> 
> _______________________________________________
> R-sig-Epi at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-epi
>