API

SimulationBasedCalibration

See documentation at: https://kiante-fernandez.github.io/SimulationBasedCalibration.jl/dev/

AbstractSBCBackend

Abstract type for SBC backends (model fitting implementations)

AbstractSBCGenerator

Abstract type for SBC generators (data generation mechanisms)

SBCDatasets

Structure to hold generated datasets and corresponding parameter values.

Fields

• variables: Array of dictionaries with true parameter values
• generated: Array of generated datasets

SBCGeneratorFunction <: AbstractSBCGenerator

A generator that uses a function to create datasets.

Fields

• func: Function that generates a single dataset
• kwargs: Keyword arguments to pass to the function

SBCGeneratorFunction(func; kwargs...)

Create a generator from a function with keyword arguments.

SBCResults

Structure to hold SBC results.

Fields

• stats: DataFrame with SBC statistics
• fits: Vector of fit objects
• errors: Vector of error objects

TuringBackend <: AbstractSBCBackend

Backend for Turing.jl models.

Fields

• model_func: Function that takes data and returns a Turing model
• sampler: MCMC sampler to use
• n_chains: Number of chains to run
• options: Additional options to pass to the Turing sampler

TuringBackend(model_func::Function; 
             sampler=Turing.NUTS(0.65),
             n_chains=4,
             kwargs...)

Create a Turing backend with the given options.

generate_datasets(generator::SBCGeneratorFunction, n_sims::Int)

Generate n_sims datasets using the given generator.

diagnostics(fit_result)

Extract diagnostics from a fit result.

diagnostics(chain::MCMCChains.Chains)

Extract diagnostics from a Turing MCMC chain.

draws_matrix(fit_result)

Convert a fit result to a draws matrix format.

draws_matrix(chain::Chains)

Convert Turing MCMCChains to a DataFrame of posterior draws.

fit(backend::AbstractSBCBackend, generated)

Fit a model using the given backend to the generated data.

fit(backend::TuringBackend, data)

Fit the Turing model to the given data.

iid_draws(backend::AbstractSBCBackend)

Return true if the backend produces independent and identically distributed draws.

iid_draws(backend::TuringBackend)

Check if the backend produces iid draws.

calculate_ranks(variables, draws)

Calculate the rank of each true parameter value within posterior draws.

compute_sbc(datasets::SBCDatasets, 
            backend::AbstractSBCBackend;
            keep_fits::Bool=true,
            thin_ranks::Int=default_thin_ranks(backend),
            ensure_num_ranks_divisor::Int=2)

Compute SBC results for the given datasets using the specified backend.

compute_statistics(fit_result, variables, thin_ranks, ensure_num_ranks_divisor)

Compute SBC statistics for a single fit.

default_thin_ranks(backend::AbstractSBCBackend)

Return the default thinning factor for rank calculation.

thin_draws(draws, thin)

Apply thinning to posterior draws.

plot_ecdf(results::SBCResults; 
          variables=nothing,
          prob=0.95)

Plot empirical CDF plots for SBC results.

plot_rank_hist(results::SBCResults; 
               variables=nothing, 
               bins=nothing,
               prob=0.95)

Plot rank histograms for SBC results.