Join the held-out predictions of a nested run onto its data

augment() returns the data a nested run was given, with each row's predictions from the outer folds that held it out. A model that never saw the row made each prediction.

Usage

# S3 method for class 'nested_results'
augment(x, ...)

Arguments

x

A nested_results from a run with save_pred = TRUE in its control.

...

Not used, and must be empty. tune's parameters argument is not offered, because each fold predicted with the parameters it selected.

Value

A tibble with the data's rows in the data's order: the outcome, then the prediction columns, whose names start with .pred, then the other data columns. For a censored outcome, the prediction columns come first. Unlike tune's augment(), it adds no .resid column.

Details

These are the outer fits' predictions, not the final model's. nested_final_fit() fits that model on all the rows, and its own augment() method predicts new data.

The outer design must hold out every data row at least once. A Monte Carlo design can miss rows, and a time-series design misses the rows before its first assessment set. For those, use collect_predictions().

On a design that holds out some row more than once, every row gets the average that collect_predictions(summarize = TRUE) gives. So a .pred_class saved beside class probabilities is recomputed from the averaged probabilities, whatever a postprocessor set. A metric computed on averaged predictions is not the nested estimate.

A row held out only by failed folds gets missing predictions, with a warning. The run needs save_pred = TRUE in its control.

See also

collect_predictions.nested_results(), compute_metrics.nested_results(), nested_final_fit(), collect_metrics.nested_results_set() for the same function on a workflow-set run

Examples

data(mtcars)

rec <- recipes::recipe(mpg ~ ., data = mtcars) |>
  recipes::step_pca(recipes::all_predictors(), num_comp = tune::tune())
wf <- workflows::workflow(rec, parsnip::linear_reg())

set.seed(1)
folds <- nested_resamples(
  mtcars,
  outside = rsample::vfold_cv(v = 2),
  inside = rsample::vfold_cv(v = 2)
)
set.seed(2)
res <- nested_tune_grid(
  wf,
  folds,
  grid = data.frame(num_comp = 1:2),
  control = tune::control_grid(save_pred = TRUE)
)

augment(res)
#> # A tibble: 32 × 12
#>      mpg .pred   cyl  disp    hp  drat    wt  qsec    vs    am  gear
#>    <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
#>  1  21    23.2     6  160    110  3.9   2.62  16.5     0     1     4
#>  2  21    23.1     6  160    110  3.9   2.88  17.0     0     1     4
#>  3  22.8  25.2     4  108     93  3.85  2.32  18.6     1     1     4
#>  4  21.4  20.0     6  258    110  3.08  3.22  19.4     1     0     3
#>  5  18.7  14.9     8  360    175  3.15  3.44  17.0     0     0     3
#>  6  18.1  21.2     6  225    105  2.76  3.46  20.2     1     0     3
#>  7  14.3  14.2     8  360    245  3.21  3.57  15.8     0     0     3
#>  8  24.4  24.6     4  147.    62  3.69  3.19  20       1     0     4
#>  9  22.8  24.1     4  141.    95  3.92  3.15  22.9     1     0     4
#> 10  19.2  22.7     6  168.   123  3.92  3.44  18.3     1     0     4
#> # ℹ 22 more rows
#> # ℹ 1 more variable: carb <dbl>