[R] gee + rcs

[Frank E Harrell Jr]

Heymans, M.W. wrote:
> Hi all,
>  
> I have fitted a gee model with the gee package and included restricted cubic spline functions. Here is the model:
>  
> chol.g <- gee(SKIN ~ rcs(CHOLT, 3), id=ID, data=chol, family=binomial(link="logit"), corstr="exchangeable")
> 
> To extract the log odds I use:
>  
> predict.glm(chol.g, type = "link")

I wonder if such predictions are 'safe', i.e., use the original knot 
locations hidden in an attribute by rcs.

>  
> Now I want to compute the logg odds for specific CHOLT values (the dependent variable) that I want to choose myself (i.e. that are not available in the dataset). Is there a way to get the linear predictor of the gee model including all separate spline functions and related coefficients? Latex from the Design package does not work here.

You'll either need to write a wrapper function for gee like glmD is for 
glm, or use lrm in Design with intra-cluster correlation correction 
using robcov or bootcov.  However this would assume a working 
independence model rather than a compound symmetry working model.

Frank

>  
> Thanks for your help!
>  
> kind regards, 
>  
> Martijn W Heymans
> 
> Faculty of Earth and Life Sciences
> 
> Institute of Health Sciences
> 
> Department of Methodology and Applied Biostatistics
> VU University 
> 
> Amsterdam, the Netherlands
> 
> 	[[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.
> 


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