应用ARIMA模型和圆分布分析病毒性脑炎病例数动态规律

目的探讨构建并应用自回归求和移动平均模型结合圆分布法分析病毒性脑炎病例数动态规律的时间序列模型方法。方法建立2001年1月至2011年12月病毒性脑炎逐月病例数的数据库。应用圆分布法分析发病高峰。建立2001年1月至2010年12月自回归求和移动平均模型的季节乘积模型ARIMA（声，d，q）×（P，D，Q）s，并利用模型预测2011年各月的病毒性脑炎病例数。结果圆分布法分析显示各年度病毒性脑炎病例数多发生于7月份，病例数高峰在7月28Et，具有明显的集中趋势（P〈0．001）；建立ARIMA（1，1，1）×（0，1，1）。。为最优模型，该模型Ljung—BoxQstatistic值为16．628，P值为0．342，NormalizedBIC为6．590。结论ARIMA乘积模型结合圆分布法是病毒性脑炎进行时间序列分析的重要方法；应用该方法对病毒性脑炎流行趋势及病例数进行预测，为卫生资源合理分配，采取开展健康教育和干预措施提供科学依据。

Dynamic analysis of patients with viral encephalitis by using ARIMA model combined with circle distribution

Objective To dynamically analyze changes of number of patients with viral encephalitis by establis- hing and applying an autoregressive integrated moving average model （ARIMA） combined with circle distri- bution. Methods Database of monthly cases of viral encephalitis from Jan. 2001 to Dec. 2011 was estab lished. Circle distribution analysis was used to verify the peak season. The multiple seasonal autoregressive in- tegrated moving average model , based on model identification was also established and was used to forecast the monthly numbers of cases for 2011. Results Circle distribution analysis showed that most cases appeared in July where it peaked on Jul. 28 （P〈0. 001） . The model was the optimal one with Ljung Box Q sta- tistic value of 16. 628 （P=0. 342） and Normalized BIC value of 6. 590. Conclusions ARIMA model com- bined with circle distribution analysis is an important method for the analysis of number of patients with viral encephalitis appeared in each month, which also can be used to predict the epidemic trend of the disease.