摘要
目的探索带有影响因素的疾病指数时间序列建模方法。方法采用黄瓜霜霉病病情指数时间序列从方法学的角度进行预测方法研究,将主成分回归模型和自回归条件异方差模型结合起来,建立AR(2)-EGARCH(0,2)模型。结果AR(2)-EGARCH(0,2)模型用应变量的过去值、过去误差和自变量的当前值、过去值的线性组合来预测病情,克服了主成分回归模型误差项不独立或存在异方差的缺点,模型取得了较好的预测效果。结论AR(2)-EGARCH(0,2)模型为本研究获得的预测效果较好的时间序列模型,适合于类似时间序列数据的结果预测。
Objective To explore the methods of disease index time series models with influencing factors. Methods Using the time series of cucumber downy mildew disease, to explore the forecasting method from the angle of methodology. The principal component regression model and autoregression conditional heteroskedastic process（ARCH） model were combined to set up AR（2） - EGARCH（0, 2）model. Results The combined linearity with the past values of dependent variables, past values of error and the present and past values of independent variables were used to forecast the disease, which can avoid the dependence of error item and the existence of heteroskectasticity of principal component regression model. Conclusion The established AR（2） - EGARCH（0, 2）model is a time series model showed satisfactory forecasting ability in this study and is suitable for the forecasting of similar time series.
出处
《中国卫生统计》
CSCD
北大核心
2006年第6期482-485,共4页
Chinese Journal of Health Statistics
基金
上海市科委科技攻关计划(03DZ19314)
