[prev in list] [next in list] [prev in thread] [next in thread] 

List:       r-devel
Subject:    Re: [Rd] makepredictcall
From:       Ben Bolker <bbolker () gmail ! com>
Date:       2022-08-13 16:52:21
Message-ID: e23ade4f-1acb-480e-02b9-9a441b686563 () gmail ! com
[Download RAW message or body]

   Well, going to git blame (too lazy to remember how to do it in svn) 
finds this commit

https://github.com/wch/r-source/commit/4c800c6f3dd61a07c763af744f08b6f411b1f877

with the commit message " c74663 failed for symbol [example(censboot, 
package="boot")] => be more careful"

  Going to ?boot::censboot and searching for predict() finds this fragment:

   t1 <- ns(d$thickness, df=4)
   cox <- coxph(Surv(d$time, d$status == 1) ~ t1+strata(d$ulcer))
   ind <- !duplicated(d$thickness)
   u <- d$thickness[!ind]
   eta <- cox$linear.predictors[!ind]
   sp <- smooth.spline(u, eta, df=20)
   th <- seq(from = 0.25, to = 10, by = 0.25)
   predict(sp, th)$y

which is exactly the use case that you identified, with "t1" as the 
culprit. (It's definitely confusing that "t1" is called out in the code 
comments, with no obvious referent ...)

  It has long been my opinion that the machinery of makepredictcall is 
poorly documented (but of course I haven't done my part and submitted 
documentation patches for consideration ...)

   cheers
    Ben Bolker



On 8/13/22 7:20 AM, Therneau, Terry M., Ph.D. via R-devel wrote:
> I'm trying to better understand makepredictcall.  The method for ns() starts with the
> following very cryptic comment and line of code:
>       ## check must work correctly when call is a symbol, both for quote(ns) and quote(t1):
>       if(as.character(call)[1L] == "ns" || (is.call(call) && identical(eval(call[[1L]]),
> ns))) {
> 
> The phrase "t1" appears no where else in the src/library/splines/R, and there is nothing
> relevant in stats/R either  (t1 gets used as a temp variable in 2 subfunctions).
> So, what is the purpose of this line of code?
> 
> I can see that if a user typed   zed <- ns, then used 'zed' in their call, that the right
> hand clause would then be true but not the left; but why worry about that case?
> 
> Terry T.
> 
>    PS (The help file isn't very helpful for creating such a method BTW.  It tells why you
> should make one, and what happens if you get it right, but nothing on how.)
> 

-- 
Dr. Benjamin Bolker
Professor, Mathematics & Statistics and Biology, McMaster University
Director, School of Computational Science and Engineering
Graduate chair, Mathematics & Statistics

______________________________________________
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
[prev in list] [next in list] [prev in thread] [next in thread]