Analysis and Generalisation of a Multivariate Exponential Smoothing Model

The multivariate exponential smoothing model of Enns, Machak, Spivey and Wrobleski is examined and it is found that its structure is such that it can be estimated by using techniques designed for a univariate exponential smoothing model. Similarly forecasts can be made using algorithms for the univariate model. The model can therefore be handled very easily. A more general univariate time series model, which can include polynomial trends and seasonal factors, is then set up and a multivariate generalisation, analogous to the multivariate exponential smoothing model, is introduced. It is shown that this model can also be handled using algorithms designed for the univariate case.