Introduction

RegimeChange is an R package for detecting structural breaks and regime changes in time series data. It provides a unified interface to both frequentist and Bayesian methods, with particular emphasis on robust performance in challenging scenarios including low signal-to-noise ratios, subtle variance changes, autocorrelated data, and heavy-tailed distributions.

Philosophical Foundations

The Ontological Problem

Regime change detection addresses a fundamental philosophical question:

How do we distinguish a real change from the inherent variability of the world?

Every observable system exhibits fluctuations. The central question is: is this observed fluctuation "noise" (random variation within the same regime) or "signal" (evidence that the system transitioned to a qualitatively different state)?

Conceptual Structure

Duality of States

We assume the world can be described through "regimes" or "states" characterized by parameters