Abstrakt
Changepoint detection is an important topic in statistical modeling. Models suggested in the past do not have to be valid any longer because of sudden economical or political shocks, natural disasters etc.
When the change in the underlying model is not discovered it may lead to the wrong forecasts, and the global accuracy of such models decays. It is therefore helpful to decide whether the change occurs in the model in the sense it is statistically significant or not.
We propose some of the tests statistics which are helpful to detect a change in parameters of the p-th order vector autoregressive process, VAR(p). The test statistics are based on the concept of maximum likelihood and limiting properties are derived using invariance principle.