Changelog
0.3.2 (2026)
Added result classes:
ConfintResult,BootstrapResult,SignificanceTestResultwith__repr__andsummary()methods.Added
SW2023Model.bootstrap()method (wrapsbootstrap_sw()).Added diagnostic plot methods:
SW2023Model.plot_efficiency(),SW2023Model.plot_frontier(),SW2023Model.plot_diagnostics().Added
plot_residuals()andplot_diagnostics()to visualization module.Added human-readable
__repr__output forSW2023ModelandPanelSW2023.Updated panel bootstrap compatibility with the panel model bandwidth argument and result dataframe output.
Bootstrap functions now accept
seedargument for reproducibility.test_r3_significancedefaultBincreased to 999.pyproject.toml: SPDX license expression,MANIFEST.inadded.Updated online documentation and replication instructions for the self-contained replication archive, revised source tarball,
requirements.txtinstallation workflow, Python 3.13 checks, manuscript figure generation, manuscript-scale full Monte Carlo table validation, and fresh Monte Carlo output generation without archived result dependencies.
0.3.1 (2025)
Initial public release.
Cross-sectional model (
SW2023Model).4-component panel model (
PanelSW2023).LOO-CV and Silverman bandwidth selection.
Pairs bootstrap CI (
bootstrap_sw(),bootstrap_panel()).Wild bootstrap significance test (
test_r3_significance()).Stata 16.1+ integration.