Detecting Change in Mean of Time Series: Likelihood Ratio Test

Keywords: Autoregressive Process; Change-point Model; Traffic Accident Data

Abstract: To detect a change in the mean of equally spaced time series data, unobservable errors in the data are modeled by an autoregressive process of order one. By employing the likelihood ratio test, a statistical method is proposed for detecting the change-point. Several examples including the Illinois traffic accident data are used to illustrate the proposed test.