search

bgreenwell / investr

Inverse estimation in R
22 stars 6 forks source link

Looks for data in global environment #42

Open stanstrup opened 5 years ago

stanstrup commented 5 years ago

This works:

mod <- glm(cbind(y, n-y) ~ ldose, data = beetle, 
           family = binomial(link = "cloglog"))

invest(mod, y0 = 0.5)

While this does not:

test_fun <- function(){
dat <- beetle
glm(cbind(y, n-y) ~ ldose, data = dat, 
           family = binomial(link = "cloglog"))
}

mod <- test_fun()
invest(mod, y0 = 0.5)

It fails with:

Error in eval(object$call$data, envir = parent.frame()) : 
  object 'dat' not found

It seems like it is looking in the global env for the object that was used to generate the model. This is problematic if you generate the model in functions or create the model on subset data that you don't keep around.

Instead I believe it should use mod$data if it needs to original data.