A Brief Overview of the Regime Shift Detection Methods

My Take

The most persistent challenge in time series analysis is correcting adequately for changes in trend. Mean stationarity is a core assumption of almost all time series forecasting algorithms, but real world data is rarely so perfect. Thankfully, there is a well developed corpus of literature in the climate research world dedicated to addressing just this issue. The terms of art in that domain: “regime shift” and “discontinuous inhomogeneity”

Their Take

Several approaches can be distinguished:

1. Parametric methods, such as the classical t-test. The methods require an assumptionabout the probability distribution of the data;

2. Non-parametric methods, such as the Mann–Whitney U-test, Wilcoxon rank sum,or Mann-Kendall test. No assumption about the probability distribution is required;

3. Curve-fitting methods;

4. Bayesian analysis and its variations, such as the Markov chain Monte Carlo method;

5. Regression-based methods;

6. Cumulative sum (CUSUM) methods;

7. Sequential methods

https://www.beringclimate.noaa.gov/regimes/rodionov_overview.pdf