Package index • multisiteDGP

Package

multisiteDGP multisiteDGP-package: multisiteDGP: Data-generating processes for multisite trial simulations

Front doors

User-facing simulation entry points. Use sim_multisite() when you can specify site sizes; use sim_meta() when you can specify information-scale targets directly.

sim_multisite(): Simulate a multisite trial data-generating process
sim_meta(): Simulate a direct-precision meta-analysis data-generating process
design_grid(): Create a Cartesian grid of multisite design cells for scenario sweeps

Design objects and data classes

Immutable design records, validators, predicates, and data coercion. Construct a design once with multisitedgp_design() when you want to reuse it across multiple simulation calls or a design_grid() sweep.

multisitedgp_design() print(<multisitedgp_design>) format(<multisitedgp_design>): Construct an immutable multisite simulation design
validate_multisitedgp_design(): Validate a multisite simulation design
update_multisitedgp_design(): Functionally update a multisite simulation design
is_multisitedgp_design(): Test whether an object is a multisite simulation design
is_multisitedgp_data(): Test for multisite simulation data objects
as_tibble(<multisitedgp_data>) `[`(<multisitedgp_data>): Coerce multisite simulation data to a plain tibble
print(<multisitedgp_data>): Print method for multisite simulation data
summary(<multisitedgp_data>): Summary method for multisite simulation diagnostics

Effect distributions

Generate the latent site-level effects at the heart of a multisite simulation. Eight built-in shapes — Gaussian, Student-t, skew normal, asymmetric Laplace, two-component mixture, point-mass slab, user callback, and a Dirichlet-process- mixture bridge — share a unit-variance convention so heterogeneity targets mean the same thing across shapes.

gen_effects(): Generate latent site effects
gen_effects_gaussian(): Generate Gaussian latent site effects
gen_effects_studentt(): Generate Student-t latent site effects
gen_effects_skewn(): Generate skew-normal latent site effects
gen_effects_ald(): Generate asymmetric Laplace latent site effects
gen_effects_mixture(): Generate two-component Gaussian mixture latent site effects
gen_effects_pmslab(): Generate point-mass-plus-slab latent site effects
gen_effects_user(): Generate user-supplied latent site effects
gen_effects_dpm(): Generate Dirichlet-process-mixture latent site effects

Site sizes and standard errors

Set how precisely each site is measured. Choose the site-size-driven model (sample sizes plus within-site variance) or the direct-precision model (information-scale targets).

gen_site_sizes(): Generate site sizes and sampling variances
gen_se_direct(): Generate direct sampling variances from informativeness and heterogeneity targets

Precision dependence

Inject intentional dependence between latent effects and site-level precision. Three injection methods — rank hill-climb, Gaussian copula, and a hybrid scheme — hit a target effect-precision correlation without distorting either margin.

align_rank_corr(): Align precision ranks to a target Spearman correlation
align_copula_corr(): Align precision ranks using a Gaussian copula pairing
align_hybrid_corr(): Align precision ranks using hybrid copula initialization and rank polishing

Observations

Observed site-level estimates from latent effects and precision — the final layer of the data-generating process.

gen_observations(): Generate observed site-level effect estimates

Diagnostics

Scenario feasibility, dependence verification, distributional fit, and shrinkage diagnostics — the four-question rubric for verifying a design behaves as intended before you commit to a long simulation run.

compute_kappa(): Compute the Neyman sampling-variance constant kappa
compute_I(): Compute realized informativeness
compute_shrinkage(): Compute per-site empirical-Bayes shrinkage proportions
informativeness(): Compute realized informativeness from data or a sampling-variance vector
mean_shrinkage(): Compute mean empirical-Bayes shrinkage
feasibility_index(): Compute the additive feasibility index (Efron or Morris)
default_thresholds(): Default thresholds for diagnostic-grid audits
heterogeneity_ratio(): Compute realized standard-error heterogeneity ratio
realized_rank_corr(): Compute realized residual-scale Spearman rank correlation
realized_rank_corr_marginal(): Compute realized marginal-scale Spearman rank correlation
bhattacharyya_coef(): Compute Bhattacharyya coefficient between empirical distributions
ks_distance(): Compute Kolmogorov-Smirnov distance between empirical distributions
scenario_audit(): Audit a grid of multisite design cells against quality-gate thresholds

Presets

Defensible starting designs for common multisite and meta-analysis questions. Each preset locks defensible parameter values and cites its source paper so a reviewer can defend the choice.

preset_education_small() preset_education_modest() preset_education_substantial() preset_jebs_paper() preset_jebs_strict() preset_walters_2024() preset_twin_towers() preset_meta_modest() preset_small_area_estimation(): Preset simulation designs — defensible starting points for common multisite scenarios

Output adapters

Coercions for downstream analysis packages — 'metafor', 'baggr', 'multisitepower' — and tidy workflows.

as_metafor() as_baggr() as_multisitepower(): Adapt multisiteDGP output for downstream meta-analysis and power packages

Visualization

Caterpillar / forest, funnel, and dependence-scatter plots for diagnosing a simulation.

plot_effects() plot_funnel() plot_dependence(): Visualization helpers for multisiteDGP simulations

Reproducibility

Stable hashes and provenance strings for simulation artifacts, so you can rerun and audit results months later.

canonical_hash(): Canonical hash for cross-machine reproducibility checks
provenance_string(): One-line human-readable provenance string