Estimation and Forecast Performance of a Multivariate Time Series Model of Sales
开发了一种名为“似不相关自回归移动平均模型”（SURARMA）的多元时间序列模型，用于预测产品在四个州的单位销量。与单变量模型相比，该模型在参数估计效率和预测表现上有显著提升，适用于涉及多个子单元的市场预测问题。
A unique form of a multivariate time series model—a “seemingly unrelated autoregressive moving average” model (SURARMA)—is developed in the context of forecasting unit sales of a product in four states. Data from an anonymous firm are used to test the appropriateness of the model and are found to conform to the model's constraints. The model provides substantial improvement in parameter estimation efficiency and forecast performance in comparison with individual state univariate models. SURARMA is potentially relevant to many market forecasting problems involving multiple constituent time series subunits such as states, regions, or products from a product line.