v0.2.0

New Features

Regression Models

• LmDynamic: Dynamic linear model for time-varying coefficients
• AID (Automatic Identification of Demand): Demand pattern classification based on Kolassa (2025)
  • ps.aid() - Classify demand patterns (regular/intermittent) with distribution selection
  • ps.aid_anomalies() - Per-row anomaly flags for stockouts, new products, obsolete products
• Formula syntax: R-style formulas with poly(), interactions (*), and I() transforms
• Prediction functions: *_predict() for all regression models with confidence/prediction intervals
• Summary functions: *_summary() for tidy coefficient output (term, estimate, std_error, statistic, p_value)

Statistical Tests

• Model classes: TTestInd, TTestPaired, MannWhitneyU, ShapiroWilk, KruskalWallis, etc.
• Full R API parity via anofox-statistics v0.3

Examples

• Added runnable examples with sample data:
  • 01_ols_regression.py - Basic OLS, predictions, formula syntax
  • 02_grouped_regression.py - Per-group regression with group_by and over
  • 03_glm_models.py - Logistic and Poisson regression
  • 04_statistical_tests.py - T-tests, Mann-Whitney, Shapiro-Wilk
  • 05_demand_classification.py - AID demand pattern classification
• Performance benchmarks for 1M groups

Improvements

• Unique prediction column names with model-type prefixes (ols_prediction, ridge_prediction, etc.)
• Updated anofox-regression to v0.4.0 (L-BFGS for Elastic Net)
• Updated anofox-statistics to v0.3 with full R API parity
• 99% test coverage
• CI improvements: coverage reporting, macOS runner update

Documentation

• Comprehensive API reference in docs/API_REFERENCE.md
• Links to anofox-regression and anofox-statistics in README

Dependencies

• anofox-regression v0.4.0
• anofox-statistics v0.3