Skip to content

scoping problem when calling lm(precomputed formula, weights) from function (PR#11540)

3 messages · Tim Hesterberg, Gabor Grothendieck, Duncan Murdoch

Original
Tim Hesterberg
# 
I've run into a scoping problem in R.
I'm calling a function that
   * creates a formula
   * calculates a weight vector
   * calls lm with that formula and weights
This fails.

Here's a simplified reproduce example:
# f works, g doesn't, h is a workaround
rm(w)
data <- data.frame(y=runif(20), x=runif(20), z=runif(20))
f <- function(k){
  w <- data$z^k
  coef(lm(y~x, data = data, weights = w))
}
g <- function(k){
  w <- data$z^k
  Formula <- parse(text="y~x")[[1]]
  coef(lm(Formula, data = data, weights = w))
}
h <- function(k){
  w <- data$z^k
  Formula <- parse(text="y~x")[[1]]
  # Following line fails due to scoping bug.  Use workaround.
  # coef(lm(Formula, data = data, weights = w))
  Call <- paste("coef(lm(", deparse(Formula), ", data=data, weights=w))")
  eval(parse(text=Call)[[1]])
}
(Intercept)           x 
  0.7452512  -0.3413998
Error in eval(expr, envir, enclos) : object "w" not found
9: eval(expr, envir, enclos)
8: eval(extras, data, env)
7: model.frame.default(formula = Formula, data = data, weights = w, 
       drop.unused.levels = TRUE)
6: model.frame(formula = Formula, data = data, weights = w, drop.unused.levels = TRUE)
5: eval(expr, envir, enclos)
4: eval(mf, parent.frame())
3: lm(Formula, data = data, weights = w)
2: coef(lm(Formula, data = data, weights = w))
1: g(2)
(Intercept)           x 
  0.7452512  -0.3413998 



--please do not edit the information below--

Version:
 platform = i486-pc-linux-gnu
 arch = i486
 os = linux-gnu
 system = i486, linux-gnu
 status = 
 major = 2
 minor = 7.0
 year = 2008
 month = 04
 day = 22
 svn rev = 45424
 language = R
 version.string = R version 2.7.0 (2008-04-22)

Locale:
LC_CTYPE=en_US;LC_NUMERIC=C;LC_TIME=C;LC_COLLATE=C;LC_MONETARY=C;LC_MESSAGES=en_US;LC_PAPER=en_US;LC_NAME=C;LC_ADDRESS=C;LC_TELEPHONE=C;LC_MEASUREMENT=en_US;LC_IDENTIFICATION=C

Search Path:
 .GlobalEnv, package:stats, package:graphics, package:grDevices, package:utils, package:datasets, package:showStructure, package:Rcode, package:splus2R, package:methods, Autoloads, package:base
Gabor Grothendieck
# 
Try replacing the last line in g with:

lmout <- do.call(lm, list(Formula, data = data, weights = w))
coef(lmout)

or replace w with environment()$w thereby explicitly telling it where to look.
On Fri, May 30, 2008 at 11:40 AM, <rocket at google.com> wrote:
Duncan Murdoch
#
On 5/30/2008 11:40 AM, rocket at google.com wrote:
No, in your use of it.
... incorrectly.
If this were simply

Formula <- y ~ x

it would work, because then the formula would get the evaluation 
environment attached.  I assume your real case builds up a formula in 
text and then parses it, in which case you'll have to evaluate the 
formula to get the environment set:

 > g
function (k)
{
     w <- data$z^k
     Formula <- eval(parse(text = "y ~ x"))
     coef(lm(Formula, data = data, weights = w))
}
 > g(2)
(Intercept)           x
   0.7777419  -0.7023185

In more complicated cases, you may need to explicitly set the 
environment of the constructed formula, using e.g.

environment(Formula) <- environment()

Duncan Murdoch