26 April 2025
Multiple Changepoint Detection for Non-Gaussian Time Series
A Flexible Approach for Identifying Changes in Complex Data
Proportion of At Bats resulting in a Home Run (HR/AB) for each Major League Baseball season from 1920 to 2023 with the optimal changepoint segmentations. Paper includes a hurricane count example as well that will be further analyzed in future work.
The Science
Detecting changes in time series data is crucial in various fields, such as finance, economics, and environmental science. However, most existing methods assume that the data follows a normal distribution, which is often not the case. Non-Gaussian data, such as count data or data with outliers, poses a significant challenge for changepoint detection.
Approach and Findings:
This study proposes a flexible approach for detecting multiple changepoints in non-Gaussian time series data. The method uses a transformation that enables it to work with different marginal distributions, allowing it to handle various types of data. The authors demonstrate the effectiveness of their method through simulations and apply it to real-world examples. The results show that the proposed method can accurately detect multiple changepoints in non-Gaussian data.
Significance and Impact:
The proposed method has significant implications for various fields, as it provides a general framework for changepoint detection that can be applied to a wide range of data types and distributions. This flexibility enables researchers and practitioners to identify changes in complex data, leading to better decision-making and more accurate predictions. The method’s ability to handle non-Gaussian data makes it particularly useful for applications where data is often non-normal, such as finance, environmental science, and public health.
Contacts (BER PM)
Renu Joseph
SC-23.1
renu.joseph@science.doe.gov (301-903-9237)
(PI Contact)
Michael Wehner
Lawrence Berkeley National Laboratory
mfwehner@lbl.gov (510-495-2527)
Funding
· This research was partially supported by the U.S. Department of Energy, Office of Science, Office of Biological and Environmental Research, Climate and Environmental Sciences Division, Regional and Global Climate Modeling Program, under award number DEAC02-05CH11231.
Publications
Robert Lund, Thomas J. Fisher, Norou Diawara, Michael Wehner (2025) Multiple Changepoint Detection for Non-Gaussian Time Series. Journal of Time Series Analysis. https://doi.org/10.1111/jtsa.12833