Score test statistic for change-point detection in AR time series with dependent errors

Abstrakt

We study the behavior of a test statistics for detecting changes in the parameters of autoregressive time series when the assumption of i.i.d. white noise is violated and replaced with the assumption of having martingal difference sequence. We present a simularion study which shows the asymptotic behavior and the power of this test statistics.

Klíčová slova

• Change-point detection
• invariance principle
• autoregressive time series