Change-Point Problems

30 Sep 2012 11:31

Suppose you have a time series which has some (stochastic) property you're interested in, say its expected value or its variance. You think this is usually constant, but that if it does change, it does so abruptly. You would like to know if and when it changes, and perhaps to localize the time when it did. You now have a change-point problem.

See also: Time Series; Non-Stationary Forecasting; Filtering and State Estimation