topepo/user2018

Slides and code for the 2018 useR! tutorial "Recipes for Data Processing"

user2018's Issues

Error: "Evaluation error: No tidyselect variables were registered", when "assess_prediction" function is used

Hi,
I am facing a problem running the "assess_prediction" function from slide 41. It gives the following error when I map it to the cv_splits object

Error in mutate_impl(.data, dots) : 
  Evaluation error: No tidyselect variables were registered
Call `rlang::last_error()` to see a backtrace.
In addition: Warning message:
Please use `new_data` instead of `newdata` with `bake`. 
In recipes versions >= 0.1.4, this will cause an error.  

When I made a slight modification to the function by removing everything() argument from bake() as shown below

assess_predictions <- function(split_obj, rec_obj, mod_obj) {
  raw_data <- assessment(split_obj)
  proc_x <- bake(rec_obj, newdata = raw_data, all_predictors())
  # Now save _all_ of the columns and add predictions. 
  bake(rec_obj, newdata = raw_data) %>%
    mutate(
      .fitted = predict(mod_obj, newdata = proc_x),
      .resid = Sale_Price - .fitted,  # Sale_Price is already logged by the recipe
      # Save the original row number of the data
      .row = as.integer(split_obj, data = "assessment")
    )
}

the function worked fine!

Would you please help me interpret the error and why it only worked after removing everything() ?

Thanks