inject_uncertainty: Inject imputation uncertainty into predictions

In VIM: Visualization and Imputation of Missing Values

Inject imputation uncertainty into predictions

Description

Adds stochastic noise to point predictions to properly reflect imputation uncertainty. Called after model prediction, before value assignment.

Usage

inject_uncertainty(
  preds,
  method = "none",
  scale = NULL,
  residuals = NULL,
  y_obs = NULL,
  score_obs = NULL,
  score_miss = NULL,
  pmm_k = 5L,
  pmm_k_method = "random",
  X_obs = NULL,
  X_miss = NULL
)

Arguments

preds	Numeric vector of predicted values for missing observations
method	One of "none", "normalerror", "resid", "pmm", "midastouch"
scale	Scale estimate (sigma hat) from model. Required for "normalerror".
residuals	Training residuals. Required for "resid".
y_obs	Observed values of target variable. Required for "pmm", "midastouch".
score_obs	Model scores for observed rows. Required for "pmm".
score_miss	Model scores for missing rows. Required for "pmm".
pmm_k	Number of donors for pmm/midastouch (default 5).
pmm_k_method	Aggregation for pmm when k > 1 (default "random").
X_obs	Predictor matrix for observed rows. Required for "midastouch".
X_miss	Predictor matrix for missing rows. Required for "midastouch".

Value

Numeric vector of adjusted predictions (same length as preds)

VIM documentation built on Sept. 2, 2026, 5:07 p.m.