v0.4.0 - FittedLikelihoodModel API

Breaking Changes

This release introduces a new statsmodels-style API for fitted models.

Old API (deprecated)

mle, iterations = model.mle(data, init)
se = model.se(mle, data)

New API

fit = model.fit(data, init)
fit.params['lambda']   # MLE estimates
fit.se['lambda']       # Standard errors
fit.llf                # Log-likelihood
fit.aic, fit.bic       # Information criteria
fit.conf_int()         # Confidence intervals
fit.wald_test(null)    # Hypothesis testing
fit.summary()          # Formatted results table

New Features

• FittedLikelihoodModel class with lazy-computed properties
• Properties: params, se/bse, llf, aic, bic, nobs, df_model, converged
• Methods: conf_int(), cov_params(), wald_test(), score_test(), lr_test(), summary()
• Works with both LikelihoodModel and ContributionModel

Documentation

• Updated all tutorials and reference docs for new API
• Live docs: https://queelius.github.io/symlik/

Full Changelog

v0.3.0...v0.4.0

symlik v0.2.0

Features

Thesis Reproduction

• Added examples/thesis_reproduction.py demonstrating 5-component Weibull series system
• Implements Bernoulli masking model satisfying C1/C2/C3 conditions
• Reproduces simulation study from "Reliability Estimation in Series Systems" (Towell, 2023)

Rerum DSL Integration

• Added examples/rerum_integration.py showing DSL-based likelihood modeling
• Added examples/likelihood.rules with reusable rule definitions
• Human-readable format for expressing distribution contributions

Series Systems

• build_weibull_series_contributions() factory function
• Support for known cause, masked cause, and censored observations
• Mixed component types (exponential + Weibull)

Documentation

• New tutorial: Thesis Reproduction
• Comprehensive API reference
• Updated README with all new features

Installation

pip install symlik

Links

• Documentation: https://queelius.github.io/symlik/
• Repository: https://github.com/queelius/symlik