R help - Dec 2004 - subsetting within a function

hi
I am trying to write a function around a glm or similar function, in which
I can pass the subsetting constraint as an argument to the function, but I
am having trouble.

the following commands (if I wanted SEX==0 to be my subset) in the global
environment work fine:

subexpr <- expression(SEX==0)
subtest <- with(mydata, eval(subexpr))
test.glm <- glm(y~x1+x2, data=mydata, family=binomial, subset=subtest)

And so my attempt to wrap a simple function around this looks like:
test.fun <- function(formula, mydata, sub=NULL){
  subs <- with(mydata, eval(sub))
  fit.glm <- glm(formula=formula, data=mydata, family=binomial, subset=subs)
  return(fit.glm)
}

But when I tested it out with
test <- test.fun(y~x1+x2, mydata=testdata, sub=expression(SEX==0))

I get:
Error in "[.data.frame"(structure(list(N_ASTHMA = as.integer(c(0, 0,
0,  :
        invalid subscript type
I'm guessing that it's looking in the global environment for
subs, but I'm not sure
how to fix this. Any help would be greatly appreciated.
--
Pamela A McCaskie
BSc(Mathematical Sciences)(Hons)

Western Australian Institute for Medical Research
University of Western Australia
SCGH Campus
Ground Floor, B Block
QE-II Medical Centre
Hospital Avenue, Nedlands
Western Australia  6009
AUSTRALIA
Email:        pmccask at cyllene.uwa.edu.au
Phone:        +61-8-9346 1612
Mob:          0417 926 607

On Thu, 30 Dec 2004, Pamela McCaskie wrote:
> And so my attempt to wrap a simple function around this looks like:
> test.fun <- function(formula, mydata, sub=NULL){
>  subs <- with(mydata, eval(sub))
>  fit.glm <- glm(formula=formula, data=mydata, family=binomial,
subset=subs)
>  return(fit.glm)
> }
>
> But when I tested it out with
> test <- test.fun(y~x1+x2, mydata=testdata, sub=expression(SEX==0))
>
> I get:
> Error in "[.data.frame"(structure(list(N_ASTHMA = as.integer(c(0,
0, 0,  :
>        invalid subscript type
I get a different error: it may be that you have an object called `subs` 
in the global environment
> I'm guessing that it's looking in the global environment for
> subs,
More precisely, it is looking in environment where `formula` was created, 
which happens to be the global environment.

This is the sort of thing that happens with the fitting functions because 
they go to such lengths to break the basic scoping of R.


You probably have to substitute() the evaluated subset into the glm call.

   fit.glm <- eval(substitute(glm(formula=formula, data=mydata,
      family=binomial, subset=subset),list(subset=subs)))


 	-thomas