Maximum likelihood estimation of a change point for Poisson distributed data

What is it about?

Our objective is to analyze non-stationary time series data. We assume that the underlying data follows the Poisson distribution and at some unknown point of time the parameters have changed. Our work not only aims to detect these changes, but also to comprehend their causes, estimate the points of time where changes occurred, and derive the distributions of these changes. Initially, we address the detection problem to determine potential shifts in time series characteristics. Subsequently, we proceed with estimating the unknown change point and deriving confidence intervals. This approach is applied to evaluate the California public safety realignment, female deaths in the U.S. due to stomach cancer and coal mining disasters in England.

Featured Image