Influential observations in bivariate VARMA models

Abstract

This article proposes a diagnostic procedure for the detection of influential observations in bivariate time series. To obtain a diagnostic measure with a reference distribution, a transformation of the bivariate VARMA model is needed. The mechanism generating outliers in a bivariate VARMA model is more complex than in the univariate case so that a strategy for the identification of the type of outliers is suggested.

The suggested diagnostic procedure is tested by simulating different types of VARMA models. The effectiveness of the method is demonstrated by analyzing real economic data.