Package index • metadid

Model fitting

Fit the hierarchical Bayesian meta-analysis model.

meta_did(): Fit a Bayesian meta-analysis model across study designs
meta_did_general(): Fit a meta-analysis model with explicit control over design assumptions
set_priors(): Set prior distributions for the meta-analysis model

Prior distributions

Specify prior distributions for model parameters.

normal(): Specify a normal prior
cauchy(): Specify a half-Cauchy prior
gamma(): Specify a gamma prior
lkj(): Specify an LKJ prior
lognormal(): Specify a log-normal prior

Results

Summarise and inspect fitted models.

print(<meta_did_fit>): Print a meta_did_fit object
summary(<meta_did_fit>): Summarise a meta_did_fit object
tidy.meta_did_fit(): Tidy a meta_did_fit object

Posterior predictive checks

Diagnose model fit via posterior predictive comparisons.

pp_check_effects(): Posterior predictive check: study-level treatment effects
pp_check_cdf(): Posterior predictive check: CDF comparison

Sensitivity analysis

Probe the assumptions that license pooling across designs, which the data cannot check: that pre-post studies share a trend distribution with DiD studies, and that all designs estimate one effect.

sensitivity_trend(): Sensitivity of the pooled effect to trend exchangeability
plot_sensitivity_trend(): Plot a trend-exchangeability sweep
omega_pp(): Pre-post share of the information about the pooled effect
delta_star(): Breakdown value for trend exchangeability
design_agreement(): Test whether the designs agree
plot_design_agreement(): Plot design offsets

Influence diagnostics

Which studies the result rests on, and which would change it. Weight governs the pull on the pooled mean; a discordant study acts on the between-study SD instead, so the two are reported separately.

information_share(): Share of the evidence each study and design contributes
plot_information_share(): Plot the share of the evidence by study
study_influence(): Per-study influence on the pooled result
plot_study_influence(): Plot per-study influence against weight

Simulation

Simulate data from the hierarchical DiD model.

simulate_meta_did(): Simulate individual-level DiD data for a collection of studies

Design extractors

Convert simulated individual-level data to the format expected by different study designs, at either the individual or summary level.

as_individual_did(): Extract full individual-level DiD data for use with meta_did()
as_individual_pp(): Extract treatment-arm individual-level pre-post data for use with meta_did()
as_individual_rct(): Extract post-only individual-level RCT data for use with meta_did()
as_summary_did(): Summarise simulated DiD data to study-level statistics
as_summary_did_change(): Summarise simulated DiD data as change-score statistics
as_summary_pp(): Summarise simulated DiD data as pre-post (treatment arm only) statistics
as_summary_rct(): Summarise simulated DiD data as RCT-style post-only statistics

Validation

Validate data inputs before fitting.

validate_summary_data(): Validate summary-level study data
validate_individual_data(): Validate individual-level study data