Everwillow Documentation

Everwillow is a statistical inference library for high-energy physics built on JAX pytrees and optimistix optimizers. It provides tools for fitting, profiling, and hypothesis testing with reusable utilities for managing optimization state.

User Guide

• Getting Started
  • Installation
  • Example Data and Tutorials
• Quickstart
  • Fitting
  • Hypothesis Testing
• Parameter Management
  • Creating and inspecting a State
  • Updating parameters
  • Partitioning and recombining
  • Merging and splitting
  • Renaming keys
• Combining Models
  • Setup: two independent measurements
  • Aligning parameter names
  • Combined fit
  • Comparing uncertainties
• Extending Hypothesis Testing
  • Custom test statistic
  • Custom distribution
  • Custom calculator
  • Custom toy generation
• Asimov Datasets
  • What is an Asimov dataset?
  • When is it needed?
  • What happens without it?
  • How to provide Asimov data
  • Choosing mu_asimov
  • Asimov as observation proxy
• Tips
  • JIT compilation
  • 64-bit precision
• Interoperability with existing statistical tools in HEP
  • The counting model
• Features
  • Interactive fitting
  • Callbacks
  • Checkpointing
  • State visualization

Developer Guide

• Contributing to everwillow

Reference

• API Reference