Contemporaneous bivariate time series

Abstract

Bivariate autoregressive-moving average time series with diagonal parameter matrices for the autoregressive and moving average components exhibit only contemporaneous or instantaneous correlation. In practice, different lengths of each series may be available. An efficient maximum likelihood algorithm for parameter estimation is derived. The statistical efficiency of this new procedure is compared with that of the standard multivariate and univariate procedures which utilize only part of the available data.