environment of lm

-----Original Message-----
From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org] On Behalf
Of William Dunlap
Sent: Sunday, June 23, 2013 1:26 PM
To: Dave Mitchell; r-help at r-project.org
Subject: Re: [R] environment of lm

Your web site had
  WeightsBreaks <- function(){
    someData <- data.frame(x = rnorm(20, 5, 1),y =  1:20 + runif(20),z =  70:51, weight =
0.05)
    print( ls())
    ModObject <- lm('x~y + z', data = someData, weights = someData$weight)
   }
and I got the result

  > print(WeightsBreaks())

  [1] "someData"
  Error in eval(expr, envir, enclos) : object 'someData' not found
When I remove the quotes around the formula string, from
  lm('x~y + z',...)
to
  lm(x~y + z, ...)
then I get

  > print(WeightsBreaks())

  [1] "someData"

  Call:
  lm(formula = x ~ y + z, data = someData, weights = someData$weight)

  Coefficients:
  (Intercept)            y            z
      47.7740      -0.5689      -0.6028

The object 'x ~ y + z' is a character string, not a formula.  It gets converted into a formula
by as.formula by lm() or perhaps by something that lm() calls.  The 'environment of a
formula' is the environment in which the formula is created (unless you explicitly change
it) so this explains why lm() could not find your stuff.

If you are are constructing your formula string using something like paste(), convert it
yourself to a formula by calling as.formula explicitly from the environment where the
objects it uses are.

Bill Dunlap
Spotfire, TIBCO Software
wdunlap tibco.com

-----Original Message-----
From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org] On Behalf
Of Dave Mitchell
Sent: Sunday, June 23, 2013 9:32 AM
To: r-help at r-project.org
Subject: [R] environment of lm

Hello,
I'm running into what I believe is a scoping issue having to do with the
environment in which the weights argument is evaluated in lm.  From the lm
documentation "All of weights, subset and offset are evaluated in the same
way as variables in formula, that is first in data and then in the
environment of formula.".  In the code at http://pastebin.com/kqzxxicp I
have 3 functions identical to each other except for

WeightsBreaks adds "weights = someData$weight"
and
Works uses the <<- assignment operator for someData.

My interpretation of the quoted documentation above is that lm looks at the
data argument to find weights, subset and offset and then to the
environment in which the formula argument lives.  Am I missing something?
Why is it that lm only sees weights (here someData$weight) when it's in the
global environment, yet lm sees data (someData in this case) regardless of
whether it's in the global environment or just local to the function?
Thanks for your time.

Dave

	[[alternative HTML version deleted]]