Answers are grounded in this paper and cite their sources.
Methodological analysis reveals an approach for detecting level shifts in non-stationary time series, highlighting applications in economic and industrial monitoring.
Suppose that observations zsubt of a time series are available at equally spaced time intervals. The authors consider the problem of making inferences about a possible shift in level of the series associated with the occurrance of an event E at some particular time. For example, the observations might be of some economic indicator and one might suspect a change in level to occur in a particular interval because of a change in fiscal policy. Alternatively, zsubt could be the daily output of a chemical process and the event E might be a change in the supplier of raw material.
No takes yet. Share an insight, caveat, or question.
Box et al. (1965) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: