R help - Feb 2015 - Environment error in the car package Anova.lme function

When I call the Anova function on a lme object from inside a function
environment I get an error:

Error in eval(expr, envir, enclos) : object 'y' not found


However, if I call the exact same code in the global environment there is
no error. My theory is that for some reason the Anova.lme function always
searches the global environment for variables rather than starting in its
parent environment.

Since the function runs correctly on lmerMod objects, and I typically
prefer lme4 to nlme this is just of academic interest.

#########
# Steps to reproduce the error

# Load Packages
library(car)
library(lme4)
library(nlme)

# Create random data.
Y <- rnorm(50)
X <- rnorm(50)
Subject <- factor(rep(LETTERS[1:5],each=10))

# Define function to fit a MEM and run the ANOVA.
fun <- function(y,x,sub){
  #mod <- lmer(y ~ x +(1 | sub) )
  mod <- lme(y~x,random=~1|sub)
  print('lme model ran successfully.')
  Anova(mod)
  print('Anova function ran successfully.')
}

# Function produces an error message.
fun(Y,X,Subject)

# if run outside of a function, it runs cleanly.
mod <- lme(Y~X,random=~1|Subject)
Anova(mod)

mod1 <- lmer(Y ~ X +(1 | Subject) )
Anova(mod1)

	[[alternative HTML version deleted]]

On Feb 24, 2015, at 10:24 AM, William Randall Henner wrote:
> When I call the Anova function on a lme object from inside a function
> environment I get an error:
> 
> Error in eval(expr, envir, enclos) : object 'y' not found
> 
> 
> However, if I call the exact same code in the global environment there is
> no error. My theory is that for some reason the Anova.lme function always
> searches the global environment for variables rather than starting in its
> parent environment.
> 
Regression functions are generally set up to first look for the tokens/symbols
that are in the formulae first in that data argument column-names and then in
the calling environment.  It's not finding an "x" or a
"y" in either of those locations (R being case sensitive). You
_should_ try to pass a dataframe with column-names that match the tokens in the
formula argument exactly to a `data=` parameter.

> Since the function runs correctly on lmerMod objects, and I typically
> prefer lme4 to nlme this is just of academic interest.
> 
> #########
> # Steps to reproduce the error
> 
> # Load Packages
> library(car)
> library(lme4)
> library(nlme)
> 
> # Create random data.
> Y <- rnorm(50)
> X <- rnorm(50)
> Subject <- factor(rep(LETTERS[1:5],each=10))
> 
> # Define function to fit a MEM and run the ANOVA.
> fun <- function(y,x,sub){
>  #mod <- lmer(y ~ x +(1 | sub) )
>  mod <- lme(y~x,random=~1|sub)
>  print('lme model ran successfully.')
>  Anova(mod)
>  print('Anova function ran successfully.')
> }
> 
> # Function produces an error message.
> fun(Y,X,Subject)
> 
> # if run outside of a function, it runs cleanly.
> mod <- lme(Y~X,random=~1|Subject)
> Anova(mod)
> 
> mod1 <- lmer(Y ~ X +(1 | Subject) )
> Anova(mod1)
> 
> 	[[alternative HTML version deleted]]
> 
> ______________________________________________
> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
David Winsemius
Alameda, CA, USA