[R] help with step()

[Sundar Dorai-Raj]

Smith, Phil wrote:
> Folks:
>  
> I'm having trouble doing a forward variable selection using step()
>  
> First, I fit an initial model:
>  
> fit0  <-  glm ( est~1 , data=all, subset=c(n>=25) )
>  
> then I invoke step():
>  
> fit1 <- step( fit0 , scope=list(upper=est~ pcped + pchosp + pfarm
> ,lower=est~1))
>  
>  
> I get the error message:  Error in eval(expr, envir, enclos) : invalid
> 'envir' argument
>  
> I looked at the documention on step(). There is no 'envir' arguement.
>  
> Can anyone shed light on what I'm doing wrong? 
>  
> Thanks,
> Phil Smith
> CDC
> 
>  
> 

Could be your "data" argument. ?all is a base function. Try re-naming to 
something else. Here's an example:

set.seed(1)
z <- all <- data.frame(x = rnorm(10), y = rnorm(10))

glm.all <- glm(y ~ 1, data = all)
step(glm.all, list(upper = y ~ x, lower = y ~ 1))

# Start:  AIC= 32.67
# y ~ 1
#
# Error in eval(expr, envir, enclos) : invalid 'envir' argument

glm.z <- glm(y ~ 1, data = z)
step(glm.z, list(upper = y ~ x, lower = y ~ 1))

# Start:  AIC= 32.67
#  y ~ 1
#
#        Df Deviance    AIC
# <none>      10.295 32.669
# + x     1    8.834 33.139
#
# Call:  glm(formula = y ~ 1, data = z)
#
# Coefficients:
# (Intercept)
#      0.2488
#
# Degrees of Freedom: 9 Total (i.e. Null);  9 Residual
# Null Deviance:      10.29
# Residual Deviance: 10.29        AIC: 32.67