[R] Re: [S] Labels wrong with lrm

[fharrell@virginia.edu]

Dear Thomas,

Your code solves the problem.  Thank you!

To users of the Design library such as Jan Brogger
who want to be able to
use variables as the second argument to one of the
library's transformation functions (pol, rcs, lsp),
you can redefine the var.inner function in the Hmisc
library as the following until the fix is made on the
source code posted on our web page.

var.inner <- function(formula) {
  if(!inherits(formula,"formula")) formula <- attr(formula,"formula")
  if(!length(formula)) stop('no formula object found')
	if(length(formula) > 2)
		formula[[2]] <- NULL  # remove response variable
  av <- all.vars(formula)
  ## Thanks to Thomas Lumley <tlumley at u.washington.edu> 28Jul01 :
  unique(sapply(attr(terms(formula),"term.labels"),
         function(term,av)
	av[match(all.vars(parse(text=term)),av)][1],
                av=av) )
}
 
-Frank Harrell


Thomas Lumley wrote:
> 
> On Sat, 28 Jul 2001 fharrell at virginia.edu wrote:
> 
> > Dear Jan,
> >
> > Thank you very much for your excellent description of the
> > problem and the self-contained test code.  This is a
> > problem that I've been meaning to either document better
> > or solve for some time.  The root of the problem is with
> > the builtin S-Plus terms.inner function:
> >
> > > attr(terms.inner(asthma ~ pol(age,kx) + smok),'variables')
> > expression(age, kx, smok)
> >
> > You can see that terms.inner inappropriately includes kx
> > as an independent variable as it does not know that
> > the first argument to pol is the special variable.
> > When a constant replaces kx, all is well.
> >
> 
> I think this can't be fixed in general.  There is simply no way to know
> whether pol(age,kx) contains two variables (like interaction(age,kx)) or
> one variable, and if so, which is the variable and which the parameter.
> 
> The termplot() function has a function carrier.names() that guesses that
> the first argument is the only variable, which is a useful heuristic until
> you have log(0.5+x) as a term.
> 
> You can get the results of all.vars broken down by term like
> 
> > f
> I(log(0.5 + x)) ~ pol(age, kx) + ns(sbp, df) + factor(race, labels =
> races) +
>     sex
> > av<-all.vars(f)
> > sapply(attr(terms(f),"term.labels"),function(term)
> av[match(all.vars(parse(text=term)),av)])
> $"pol(age, kx)"
> [1] "age" "kx"
> 
> $"ns(sbp, df)"
> [1] "sbp" "df"
> 
> $"factor(race, labels = races)"
> [1] "race"  "races"
> 
> $sex
> [1] "sex"
> 
> (actually you only get the RHS of the formula, but that shouldn't be hard
> to fix)
> 
> You might also look at how nlme and Jim Lindsey's nonlinear models
> functions solve the question of whether a name refers to a a variable or a
> parameter.
> 
>         -thomas
> 
> Thomas Lumley                   Asst. Professor, Biostatistics
> tlumley at u.washington.edu        University of Washington, Seattle
> 
> -.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
> r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
> Send "info", "help", or "[un]subscribe"
> (in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
> _._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

-- 
Frank E Harrell Jr              Prof. of Biostatistics & Statistics
Div. of Biostatistics & Epidem. Dept. of Health Evaluation Sciences
U. Virginia School of Medicine  http://hesweb1.med.virginia.edu/biostat
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._