Default ABC discrepancy: censoring-aware weighted RMSE on the log10 scale

Detected observations contribute squared error on the log10 scale. Left-censored observations are penalized only when the predicted trajectory exceeds the observation-specific LOD.

Usage

mbmm_default_distance(
  y_obs,
  pred,
  censored = NULL,
  lod = NULL,
  w_detect = 5,
  w_censored = 1,
  floor = 0.01
)

Arguments

y_obs

Observed values on the natural scale.

pred

Predicted values on the natural scale.

censored

Logical or 0/1 vector indicating left-censored observations.

lod

Observation-specific LOD on the natural scale.

w_detect

Weight for detected-observation squared errors.

w_censored

Weight for censored-observation squared errors.

floor

Smallest positive value passed to log10.

Value

A nonnegative scalar discrepancy.