NEWS.md

In eam: Evidence Accumulation Models

eam 1.2.0

New Features

• ABI (Approximate Bayesian Inference) Module: Complete neural network-based parameter estimation workflow
• abi_train(): Train neural estimators using simulation-based inference
• abi_estimate(): Obtain point estimates from trained models
• abi_assess(): Assess trained estimator performance
• abi_sample_posterior(): Sample from posterior distribution

• Enhanced build_abi_input() with theta and Z outputs, test set support

• ABC helpers: Add abc_abc() and abc_cv() wrappers for ABC fitting and cross-validation

• Posterior predictive workflows: Add abc_posterior_predictive_check(), abi_posterior_predictive_check(), and update_config_from_posterior() for teaching-oriented posterior simulation workflows

• Visualization:

• New plot_cv_recovery() methods for ABI models (eam_abi_assess and eam_abi_posterior_samples classes)
• Update posterior RT and accuracy plots to compare simulated and observed data more directly

• plot_rt() now displays simulated RTs as densities and observed RTs as histograms

• Posterior summarization: summarise_posterior_parameters() for aggregating posterior samples

Infrastructure

• Julia environment integration via init_julia_env() for neural network backend
• Add bundled Julia project files under inst/julia/env/ for ABI setup
• Add tibble dependency for improved output formatting
• Improve simulation routing, including LFM support and updated LBA routing

Documentation

• Add inst/CITATION file with pre-print citation information
• Add and expand examples for ABC wrappers, posterior predictive helpers, and simulation-config updates
• Add ABI and ABC demos and refresh tutorial materials
• Various documentation improvements and formatting fixes

eam 1.1.0

• Add build_abi_input function to create input for ABI anlysis from EAM simulation output.
• Simulation allow more than 1024 data chunks/arrow partitions. Now, it depends on the hard limit of the arrow library and the file system.
• Fix summarise_by() to handle invalid column names returned by summary functions (e.g., quantile functions returning "90%", "95%"). Now uses vctrs::vec_as_names() for proper name repair.
• By convention of ABC, change the prior of plot_posterior_parameters to the hist graph.
• By convention of ABC, change the posterior of plot_rt to reflect the median RT within each condition.

repo

• Add documentation with pkgdown.

eam 1.0.1

• Initial release of this package.

eam documentation built on March 29, 2026, 5:07 p.m.