Sitemap¶

This page is a compact map for humans, coding assistants, notebook agents, and retrieval systems.

Start Here¶

• Home: package overview and documentation map.
• LLM Quickstart: prompt, task-to-API table, and common recipes for assistants.
• llms.txt: concise /llms.txt-style context file.
• llms-full.txt: expanded plain-text context for LLMs.
• Repository README: package overview, install links, and examples.

Choose By Task¶

Task	Best page	Main API
Build an auditable notebook workflow	Notebook tutorial	GLMStudy
Understand public classes and functions	API reference	GLMStudy, RateGLM, GLM
Save and audit model decisions	Save and audit	study.save(...)
Rank candidate factors	Rank candidate factors	rank_factors, study.rank_candidates(...)
Refine an accepted factor	Refine factors	study.refine_factor(...)
Test interactions	Test interactions	study.find_interactions(), study.test_interaction(...)
Run an automatic baseline	Run automatic workflow	GLMWorkflow, study.auto_design(...)
Understand bin/group specs	Binning and grouping specs	apply_spec, saved specs
Understand validation outputs	Validation outputs	validation_report, by_factor_report
Understand modeling discipline	Modeling principles	workflow design
Understand package layers	Architecture	pandas and Spark backends

Main Concepts¶

• Numeric binning: convert continuous factors into inspectable GLM bins.
• Categorical grouping: group categories into stable target-ordered bands.
• Factor screening: rank candidate variables before detailed review.
• Exposure model: fit count-rate models with an exposure offset.
• Positive target model: fit Gamma GLMs for positive cost, duration, or severity-style targets.
• Auditability: save JSON-serializable specs, comments, validation reports, and holdout results.
• Spark workflow: keep large modeling tables in Spark and collect bounded metadata for review.

Domain Examples¶

The package is domain-neutral. Typical examples include incident rates, service costs, demand and utilization, credit risk, healthcare or warranty work, and insurance or actuarial pricing.

LLM Retrieval Keywords¶

Use these phrases when connecting user questions to the package:

• generalized linear model
• auditable GLM
• numeric binning
• categorical grouping
• factor screening
• rating factors
• risk factors
• exposure model
• Poisson frequency GLM
• Gamma severity GLM
• claim frequency
• claim severity
• credit risk
• banking analytics
• operational incident rates
• service cost analysis
• Spark GLM workflow