Change printing for `step_impute_knn()`

[SewerynGrodny]

Step_impute_knn - ignore variable that should be imputed, instead of this impute value in all possible numeric variables:
prep(base_rec)

K-nearest neighbor imputation for Sepal.Width, Petal.Width, Species [trained]

whereas it should be only Petal.Length.

Bests
Seweryn

iris = as_tibble(iris)

iris[1, 1] <- as.numeric(NA)
iris[1, 2] <- as.numeric(NA)
iris[1 ,3] <- as.numeric(NA)
iris[1, 4] <- as.numeric(NA)
iris[1, 5] <- as.numeric(NA)

set.seed(123)

iris_split <- iris %>% 
  initial_split(strata = Sepal.Length)

iris_training <- training(iris_split)
iris_testing <- testing(iris_split)

iris_rf_model <- rand_forest(
  mtry = 5,
  min_n = 5,
  trees = 500) %>%
  set_engine("ranger") %>%
  set_mode("regression")


base_rec <- recipe(Sepal.Length ~ .,
                   data = iris_training) %>% 
  step_impute_knn(Petal.Length) 

prep(base_rec)

[juliasilge]

There are a couple of things going on here that may be contributing to some confusion. When you do the data splitting, you end up with data in iris_training that doesn't have any NA values, so there is nothing that needs imputation. Try it just using the tibble that you add the missing data to, and you'll notice that it will fail because there is nothing to use for imputation.

If instead we only add two NA value, you'll notice that the correct column (Petal.Length) is being imputed:

library(tidymodels)
#> Registered S3 method overwritten by 'tune':
#>   method                   from   
#>   required_pkgs.model_spec parsnip

iris <- tibble(iris)
iris[1, 2] <- NA_real_
iris[1, 3] <- NA_real_
iris
#> # A tibble: 150 × 5
#>    Sepal.Length Sepal.Width Petal.Length Petal.Width Species
#>           <dbl>       <dbl>        <dbl>       <dbl> <fct>  
#>  1          5.1        NA           NA           0.2 setosa 
#>  2          4.9         3            1.4         0.2 setosa 
#>  3          4.7         3.2          1.3         0.2 setosa 
#>  4          4.6         3.1          1.5         0.2 setosa 
#>  5          5           3.6          1.4         0.2 setosa 
#>  6          5.4         3.9          1.7         0.4 setosa 
#>  7          4.6         3.4          1.4         0.3 setosa 
#>  8          5           3.4          1.5         0.2 setosa 
#>  9          4.4         2.9          1.4         0.2 setosa 
#> 10          4.9         3.1          1.5         0.1 setosa 
#> # … with 140 more rows

base_rec <- recipe(Sepal.Length ~ ., data = iris) %>% 
  step_impute_knn(Petal.Length) 

prep(base_rec)
#> Recipe
#> 
#> Inputs:
#> 
#>       role #variables
#>    outcome          1
#>  predictor          4
#> 
#> Training data contained 150 data points and 1 incomplete row. 
#> 
#> Operations:
#> 
#> K-nearest neighbor imputation for Sepal.Width, Petal.Width, Species [trained]
prep(base_rec) %>% bake(new_data = NULL)
#> # A tibble: 150 × 5
#>    Sepal.Width Petal.Length Petal.Width Species Sepal.Length
#>          <dbl>        <dbl>       <dbl> <fct>          <dbl>
#>  1        NA           1.44         0.2 setosa           5.1
#>  2         3           1.4          0.2 setosa           4.9
#>  3         3.2         1.3          0.2 setosa           4.7
#>  4         3.1         1.5          0.2 setosa           4.6
#>  5         3.6         1.4          0.2 setosa           5  
#>  6         3.9         1.7          0.4 setosa           5.4
#>  7         3.4         1.4          0.3 setosa           4.6
#>  8         3.4         1.5          0.2 setosa           5  
#>  9         2.9         1.4          0.2 setosa           4.4
#> 10         3.1         1.5          0.1 setosa           4.9
#> # … with 140 more rows

^{Created on 2021-10-20 by the reprex package (v2.0.1)}

The Sepal.Width is not getting imputed because we didn't ask this recipe to do that.

The printing is confusing because it is telling you the things that are possibly being used for imputation and not what is being imputed. Maybe we can fix that here:

recipes/R/impute_knn.R

Line 275 in ca9f84b

cat("K-nearest neighbor imputation for ", sep = "")

We could just change "for" to "with"? Or change which variables are printed out and do the ones that are being imputed rather than the ones that are being imputed with.

[SewerynGrodny]

Hi, thanks, for an explanation.
I like you proposition by changing message from what was used for imputation on what had been imputed.
Thanks,
PS. Should I close this issue?

[juliasilge]

No, we'll leave this open and close it when we change the printing! 👍

[github-actions]

This issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with a reprex https://reprex.tidyverse.org) and link to this issue.