【摘要】 目的应用三种统计模型预测伤寒副伤寒的发病趋势,比较其预测效果,为伤寒副伤寒的预测和防控提供科学依据。方法利用广东省2008年5月至2012年4月四年的伤寒副伤寒逐月发病资料,分别拟合季节性综合自回归滑动平均(SARIMA)模型、经傅里叶季节性调整的综合自回归滑动平均(FARIMA)模型和动态谐波回归(DHR)模型,并用前面建立的三种模型预测后续半年(2012年5月—2012年10月)的逐月发病数。结果伤寒副伤寒的发病有明显的周期性和季节特征,周期为1年,7-8月份为发病高峰期。流行强度和流行高峰出现的月份均存在一定的年度差异。三种模型拟合四年的伤寒副伤寒发病情况,其平均绝对百分比误差(MAPE)依次为:DHR模型(7.8%)<FARIMA模型(12.9%)<SARIMA模型(13.4%);三种模型预测后续半年的发病情况,其MAPE依次为:DHR模型(3.5%)<FARIMA模型(5.6%)<SARIMA模型(6.8%),其他模型评价指标结果也类似。结论三种方法均有较佳的预测效果。相对而言,DHR的预测精度更高。本研究可为常见传染病的预测提供一定的方法学参考。更多还原

【Abstract】 Objective To compare the performance of three statistical methods in forecasting the incidence of typhoid fever and paratyphoid fever. Methods Using monthly data of cases with typhoid fever and paratyphoid fever in Guangdong Province from May 2008 to April 2012,we fitted three models,including Seasonal Autoregressive Integrated Moving Average( SARIMA) model,ARIMA model applied to the seasonally adjusted data from Fourier terms( FARIMA) and Dynamic Harmonic Regression( DHR) model. Afterwards,we used the data from May 2012 to October 2012 to assess the accuracy of prediction for these three models. Results The incidence of typhoid fever and paratyphoid fever is characterized by an apparent cyclic pattern with a one- year seasonal cycle. The maximum number of cases occurred during July to August. The epidemic strength and peak differed by years. Mean Absolute Percentage Error( MAPE) of the four- year time series fitted with three methods was: DHR( 7. 8%) < FARIMA( 12. 9%) < SARIMA( 13. 4%),and MAPE of the half-a- year time series forecasted had the same trend: DHR( 3. 5%) < FARIMA( 5. 6%) < SARIMA( 6. 8%). Other index for forecasting accuracy conveyed the similar message. Conclusion All three models had satisfactory prediction capacity. DHR model is superior to FARIMA model and SARIMA model with a higher forecast precision. The methods and findings in this study may throw light on predicting the incidence of other infectious diseases.更多还原