Best Imputation Method per Column

Picks, for each variable, the method with the lowest relative squared error.

Usage

tbl_best_imputation_method(rse_df)

Arguments

rse_df

A [compare_imputation_rse()] result, or [tune_impute_numeric()] output (anything with `variable`, `method` and `rse`).

Value

A tibble with one row per variable: `variable`, `best_method`, `rse`, `runner_up`, `runner_up_rse`, `beats_null`, `n_methods`.

Details

Ties go to the method that comes first in alphabetical order; methods with a missing RSE are never chosen (a variable where all are missing gets `NA`). `beats_null` says whether the winner's RSE is below 1.

Examples

rse <- data.frame(variable = c("x", "x", "y", "y"),
                  method = c("mean", "knn", "mean", "knn"), rse = c(1, 0.4, 0.9, 1.2))
tbl_best_imputation_method(rse)
#> # A tibble: 2 × 7
#>   variable best_method   rse runner_up runner_up_rse beats_null n_methods
#>   <chr>    <chr>       <dbl> <chr>             <dbl> <lgl>          <int>
#> 1 x        knn           0.4 mean                1   TRUE               2
#> 2 y        mean          0.9 knn                 1.2 TRUE               2