Optimal online detection of parameter changes in two linear models (Q1073519)

This paper is concerned with the problem of sequential detection of certain parameter changes in the case of dependent sequences. Two cases are discussed: one is a shift in the mean of an autoregressive process, and the other is a change in the regression coefficient of a regression model with serially correlated error terms. The detection rule derived in this paper is optimal in the sense of minimizing the expected delay in detection of the change under some restriction of expected number of false signals of change.