Forecasting Using Long‐Order Autoregressive Processes: An Example Using Housing Starts
研究了一种长阶自回归模型用于预测美国月度住房开工数据，无需差分处理趋势和季节成分，预测误差方差小于Box-Jenkins ARIMA方法，适合具有强季节性和缓慢变化趋势的住房市场时间序列。
A long autoregressive (AR) modeling procedure for monthly U.S. housing starts data is considered. Neither differencing to remove the trend, nor differencing to remove the seasonal component is required in this method. The model is fitted by a Householder transformation‐Akaike AIC criterion algorithm. Forecast performance is compared to that obtained by the Box‐Jenkins ARIMA method. The prediction error variance of the long AR model method tends to be smaller than the prediction error variance of the Box‐Jenkins model method. The long AR method is well suited for housing market time‐series which are characterized by both strong seasonal and slowly changing trend components.