Function reference • debiasLLMReporting

Core Estimators

Main functions for debiased prevalence estimation

ppi_point_and_ci(): Prediction-powered mean estimator with logit CI.

ppi_pp_point_and_ci(): PPI++ estimator with optimal unconstrained lambda.

rg_point_and_ci(): Rogan-Gladen corrected estimator with finite-sample adjustment.

eif_point_and_ci(): EIF (Efficient Influence Function) estimator for binary surrogates.

mle_point_and_ci(): MLE estimator with consistent output format.

fit_misclass_mle(): Joint MLE under the binary misclassification model.

grab_theta(): Grab the theta-specific summary from a joint MLE fit.

Discrete Estimators

Estimators for discrete outcomes

eif_discrete_point_and_ci(): EIF estimator with per-category calibration (optimal for discrete)

eif_gam_discrete_point_and_ci(): EIF estimator with GAM calibration

eif_linear_discrete_point_and_ci(): EIF estimator with linear calibration (suboptimal for discrete)

eif_spline_discrete_point_and_ci(): EIF estimator with spline calibration

naive_discrete_point_and_ci(): Naive estimator: just use mean of discrete predictions

oracle_discrete_point_and_ci(): Oracle estimator (knows true \(\mu\) values)

ppi_discrete_point_and_ci(): PPI estimator for discrete predictions

ppi_pp_discrete_point_and_ci(): PPI++ estimator for discrete predictions

Continuous Estimators

Estimators for continuous outcomes

eif_continuous_point_and_ci(): EIF estimator for continuous surrogates with calibration

naive_continuous_point_and_ci(): Naive estimator (just use model predictions)

ppi_continuous_point_and_ci(): PPI estimator for continuous scores

ppi_pp_continuous_point_and_ci(): PPI++ estimator for continuous scores

Calibration Functions

Functions for fitting calibration models

fit_calibration_gam(): Fit calibration model using GAM

fit_calibration_isotonic(): Fit calibration model using isotonic regression

fit_calibration_linear(): Fit calibration model using linear regression

fit_calibration_spline(): Fit calibration model using cubic spline

Data Generation

Functions for generating simulated data

generate_discrete_dgp(): Generate discrete prediction data

Helper Functions

Utility functions

logit_ci(): Wald-style confidence interval via logit transform.

Visualization

Plotting functions for simulation results

plot_coverage(): Plot empirical coverage across simulation settings

plot_ci_width(): Plot mean confidence interval width across simulation settings

plot_bias(): Plot estimator bias across simulation settings

plot_bias_pct(): Plot percent bias across simulation settings

plot_q_bias(): Plot calibration bias for \(\hat q_0\) and \(\hat q_1\)

build_line_plot(): Generic line plot builder

generate_plot_suite(): Generate all 5 standard plots for a given data subset

get_method_color_scale(): Get the fixed color scale for methods

plot_base_theme(): Base theme for standard plots

plot_base_theme_aggregate(): Larger legend theme for aggregate plots (stacked in two rows)

format_label_ratio(): Format a labeling ratio as a percentage string