Skip to contents

Detect Outlier Samples

Source: R/qc.R
step_measure_qc_outlier.Rd

step_measure_qc_outlier() creates a specification of a recipe step that detects outlier samples using Mahalanobis distance or PCA-based methods. A new column is added indicating outlier status.

Usage

step_measure_qc_outlier(
  recipe,
  measures = NULL,
  method = c("mahalanobis", "pca"),
  threshold = 3,
  n_components = NULL,
  new_col = ".outlier",
  new_col_score = ".outlier_score",
  role = "predictor",
  trained = FALSE,
  skip = FALSE,
  id = recipes::rand_id("measure_qc_outlier")
)

Arguments

recipe

A recipe object.

measures

An optional character vector of measure column names.

method

Detection method:

  • "mahalanobis" (default): Mahalanobis distance with robust covariance

  • "pca": PCA score-based outliers (Hotelling's T^2)

threshold

Threshold for outlier detection in standard deviation units. Default is 3. Tunable via outlier_threshold().

n_components

For PCA method, number of components to use. Default is NULL (auto-select based on variance explained).

new_col

Name of the new outlier flag column. Default is ".outlier".

new_col_score

Name of the outlier score column. Default is ".outlier_score".

role

Role for new columns. Default is "predictor".

trained

Logical indicating if the step has been trained.

skip

Logical. Should the step be skipped when baking?

id

Unique step identifier.

Value

An updated recipe with the new step added.

Details

Outlier samples can arise from measurement errors, sample preparation issues, or genuine unusual samples. This step helps identify them.

Mahalanobis method: Computes the multivariate distance from each sample to the center of the distribution, accounting for correlations. Uses robust estimation of center and covariance via median and MAD.

PCA method: Projects data onto principal components and computes Hotelling's T^2 statistic. Samples with extreme scores are flagged.

Two columns are added:

  • .outlier: Logical flag

  • .outlier_score: Numeric score (higher = more extreme)

See also

Other measure-qc: step_measure_impute(), step_measure_qc_saturated(), step_measure_qc_snr()

Examples

library(recipes)

rec <- recipe(water + fat + protein ~ ., data = meats_long) |>
  update_role(id, new_role = "id") |>
  step_measure_input_long(transmittance, location = vars(channel)) |>
  step_measure_qc_outlier(threshold = 3) |>
  prep()

bake(rec, new_data = NULL)
#> # A tibble: 215 × 8
#>       id water   fat protein .measures channel     .outlier .outlier_score
#>    <int> <dbl> <dbl>   <dbl>    <meas> <list>      <lgl>             <dbl>
#>  1     1  60.5  22.5    16.7 [100 × 2] <int [100]> FALSE             0.305
#>  2     2  46    40.1    13.5 [100 × 2] <int [100]> FALSE             0.323
#>  3     3  71     8.4    20.5 [100 × 2] <int [100]> FALSE             0.626
#>  4     4  72.8   5.9    20.7 [100 × 2] <int [100]> FALSE             0.133
#>  5     5  58.3  25.5    15.5 [100 × 2] <int [100]> FALSE             0.310
#>  6     6  44    42.7    13.7 [100 × 2] <int [100]> FALSE             0.793
#>  7     7  44    42.7    13.7 [100 × 2] <int [100]> FALSE             0.731
#>  8     8  69.3  10.6    19.3 [100 × 2] <int [100]> FALSE             0.522
#>  9     9  61.4  19.9    17.7 [100 × 2] <int [100]> FALSE             1.44 
#> 10    10  61.4  19.9    17.7 [100 × 2] <int [100]> FALSE             1.78 
#> # ℹ 205 more rows