[R] rcspline.problem

[Peter Ehlers]

On 2012-07-19 07:10, Bart Ferket wrote:
> Dear professor Harrell,
>
> I probably have the same problem as Haleh Ghaem Maralani.
>
> I am using the rms package and the rcspline.plot function to assess the
> relation of a continuous predictor to the log hazard function.
>
> I would like to use the "adj" statement, for example using this test
> dataset:
>
> http://r.789695.n4.nabble.com/file/n3058505/file.csv
>
> test <-read.csv("file.csv",header=TRUE)
>
> rcspline.plot(test$factor, test$Time,model="cox", adj=cbind(test$adj1,
> test$adj2),
> xrange=c(0,3),ylim=c(-1,2),event=test$event,nk=4,knots=c(0.8,1.0,1.5,2.0),showknots=TRUE,plotcl=FALSE,statloc="none",subset=test$SEX=="2",lty=2)
>
> Then I get the following Error.
>
>           x
> -0.7860188  3.4871734 -4.6087226 -0.6761077 -0.9358280
> [1] -3686.955 -3646.681
> Error in pchisq(q, df, lower.tail, log.p) :
>    Non-numeric argument to mathematical function
>
> Could you please help us understanding how to use the adj statement?
>
> Bart Ferket, MD, dept. Epidemiology, ErasmusMC Rotterdam

[You're replying to a year-old post; it would probably have
been better to start a new thread.]

I think that there may be a bug in rcspline.plot that
may be easily fixed. Grab the code of rcspline.plot()
and edit it as follows:

1. Find the line with

       adj.df <- attr(v, "rank")

and replace it with

       adj.df <- qr(v)[["rank"]]

2. Save the function as "myrcspline.plot".

3. Fix the evironment of your new function:

    environment(myrcspline.plot) <- environment(rcspline.plot)

Now use myrcspline.plot in your analysis. Works for me, but
I'm no expert in such models and don't what side effects
this change might have.

Peter Ehlers