[R] predict()

[Ivo Shterev]

Dear Dr. Therneau,

Thank you for your response. Just to point out that we didn't experience any problems with the lm() function under R version 2.12.2 (2011-02-25):

> set.seed(123)
> testdat=data.frame(y=rexp(10),event=rep(0:1,each=5),x=rnorm(10))
> testfm=as.formula('y~x')
> 
> testfun=function(dat,fm)
+   {
+     predict(lm(fm,data=dat),newdata=dat)
+   }
> testfun(testdat,testfm)
          1           2           3           4           5           6 
 1.00414971  0.07061335  0.55381658  0.53091759  0.69310319  1.06574974 
          7           8           9          10 
-0.24405980  0.47664172  1.85449993  0.36286396 
> sessionInfo()
R version 2.12.2 (2011-02-25)
Platform: x86_64-pc-linux-gnu (64-bit)

locale:
 [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C              
 [3] LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8    
 [5] LC_MONETARY=C              LC_MESSAGES=en_US.UTF-8   
 [7] LC_PAPER=en_US.UTF-8       LC_NAME=C                 
 [9] LC_ADDRESS=C               LC_TELEPHONE=C            
[11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C       

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

loaded via a namespace (and not attached):
[1] tools_2.12.2
> 


Regards
Ivo



--- On Thu, 4/14/11, Terry Therneau <therneau at mayo.edu> wrote:

> From: Terry Therneau <therneau at mayo.edu>
> Subject: Re: predict()
> To: "Ivo Shterev" <idc318 at yahoo.com>
> Cc: r-help at r-project.org
> Date: Thursday, April 14, 2011, 2:59 PM
> --- begin included message ---
> I am experimenting with the function predict() in two
> versions of R and
> the R extension package "survival".
> 
> library(survival)
> 
> set.seed(123)
> testdat=data.frame(otime=rexp(10),event=rep(0:1,each=5),x=rnorm(10))
> testfm=as.formula('Surv(otime,event)~x')
> 
> testfun=function(dat,fm)
>   {
>    
> predict(coxph(fm,data=dat),type='lp',newdata=dat)
>   }
> 
> -- end inclusion ----
> 
>  The question was: this works under survival 2.35-8, but
> not survival
> 2.36-5
> 
> Answer: The predict and underlying model.frame functions
> for coxph were
> brought into line with lm and other models.  The major
> advantage is that
> I now deal with factors and data dependent predictors like
> ns()
> correctly.
>   You've shown a disadvantage of which I was not
> aware.  Using your
> example but replacing coxph() by lm() with otime ~x as the
> model I get a
> similar failure.  I'd like to ask a wider audience of
> R-devel since it
> is bigger than coxph.
> 
> Terry Therneau
> 
> 
>