Skip to contents

`margot_lmtp_censoring_report()` reports observed retention and, when supplied, fitted continued-observation probabilities and censoring factors. The function returns aggregate evidence only. It neither classifies censoring support nor returns a route action.

Usage

margot_lmtp_censoring_report(
  observed,
  baseline_weights = NULL,
  fitted_probabilities = NULL,
  censoring_factors = NULL,
  joint_ratios = NULL,
  wave_labels = NULL,
  policy_id = NULL,
  learner_specification = NULL,
  out_of_fold_performance = NULL,
  na_is_unobserved = TRUE
)

Arguments

observed

A logical or `0`/`1` matrix with one participant per row and one censoring transition per column. A vector represents one transition.

baseline_weights

Optional non-negative baseline design weights, one per participant. Equal weights are used when this argument is `NULL`.

fitted_probabilities

Optional matrix of fitted probabilities of continued observation, with the same dimensions as `observed`.

censoring_factors

Optional matrix of separately identified censoring density-ratio factors, with the same dimensions as `observed`.

joint_ratios

Optional matrix of joint exposure-and-censoring density ratios, with the same dimensions as `observed`.

wave_labels

Optional transition labels. Column names from `observed` are used when available.

policy_id

Optional policy identifier recorded as provenance.

learner_specification

Optional aggregate description of the registered censoring learners.

out_of_fold_performance

Optional aggregate out-of-fold performance record for the censoring learners.

na_is_unobserved

Logical; whether an `NA` observation indicator denotes loss to follow-up. The default is `TRUE`.

Value

An object of class `margot_lmtp_censoring_report` containing retention, probability, factor, zero-cause, learner, and provenance records. The computed tables contain no participant-level rows; supplied learner records must likewise be aggregate.