calibration and validation for svycoxph

[R user]

Hello,

This post is for Dr. Thomas Lumley or anybody familiar with the survey package. I am estimating a proportional hazards model with weights using svycoxph. Are there functions already built in the survey package that allow me to do validation and calibration of the model?

Thanks

[Thomas Lumley]

You don't say what you mean by validation and calibration of the model (these terms have multiple meanings) but if you mean looking at how good the prediction is, all the raw materials are available through svykm() and the predict() method for svycoxph().  There aren't any predefined functions like the ones Frank Harrell provides, though.

Incidentally, sending two messages in quick succession like this is not a good strategy for getting help.

    -thomas


On Wed, 9 Jun 2010, R user wrote:

>
> Hello,
>
> This post is for Dr. Thomas Lumley or anybody familiar with the survey
> package. I am estimating a proportional hazards model with weights using
> svycoxph. Are there functions already built in the survey package that allow
> me to do validation and calibration of the model?
>
> Thanks
> --
> View this message in context: http://r.789695.n4.nabble.com/calibration-and-validation-for-svycoxph-tp2249118p2249118.html
> Sent from the R help mailing list archive at Nabble.com.
>
> ______________________________________________
> [hidden email] 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.
>

Thomas Lumley Assoc. Professor, Biostatistics
[hidden email] University of Washington, Seattle

______________________________________________
[hidden email] 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.

[R user]

Hi Dr. Lumley,

Sorry about the succession of the 2 messages. It wasn't intended. This is my first post to the forum and I first signed up to Nabble and submitted the post and later got the message from Nabble that the post has not been accepted by the R forum and that I need separate subscription for the R forum. So I signed up for the R forum and resubmitted the post and both messages ended up being posted.

Yes, I was looking for similar functions to those in the Design package to build a calibration plot (Design, calibrate function) and to do bootstrap validation of the model (Design, validate function). I will check the options of the functions you mentioned to see if I can easily get those. Could you please advise on what measure should I use to quantify the optimism of fit (the way c-index was used for Cox models)? If you have existent code that you used for these purposes that would be great.

Many thanks.