浙江省食物中毒发生趋势的时间序列模型研究

A study on time series model of incidence trend of food poisoning in Zhejiang Province

目的建立浙江省食物中毒发生趋势的数学模型并探讨其可行性。方法应用自回归求和移动平均模型(autoregression integated moving average model,ARIMA)分析1992-2006年浙江省l5年食物中毒发生起数和人数分别建立相应的模型,用2007年食物中毒发生情况对模型的进行合理性验证,对2008年食物中毒发生情况进行预测。结果用复合季节模型ARIMA(0,0,0)×(0,1,1)4、ARIMA(1,0,0)×(0,1,1)4模型分别对浙江省食物中毒发生起数和中毒发生人数数据拟合,自回归系数差异均有统计学意义(P<0.05),白噪声残差分析显示残差序列自相关函数的Box?Ljung Q统计检验差异无统计学意义(P>0.05),残差为随机性误差;中毒起数的ARIAM模型为lnyt=lnyt-4-0.637εt-4、中毒人数的ARIAM模型为yt=yt-4+0.331yt-1-0.331yt-5-0.746εt-4;2007年实际值均在其预测值95%可信区间内,符合率达100%。结论ARIAM模型是用于拟合食物中毒发生趋势和预测的一种行之有效的统计方法。

Objective To construct the mathematical model of incidence trend of food poisoning in Zhejiang Province and to investigate its feasibility.Methods Food poisoning seasonal surveillance data from 1992 to 2006 were analyzed by using autoregressive integrated moving average model（ARIMA） and were used to establish prediction model.Information of 56 preceding seasons was used to establish the ARIMA model and data of 4 succeeding seasons were evaluated.Results Model ARIMA（0,0,0）×（0,1,1）4 and ARIMA（1,0,0）×（0,1,1）4 were from food poisoning times and food poisoning persons,respectively（P〈0.001）.White noise analysis showed that the minimum Box-Ljung Q value of autocorrelation function was not significant and the residual was randomized error.Prediction models of lnyt=lnyt-4-0.637εt-4 for food poisoning times and √yt=√yt-4＋0.331√yt-1-0.331√yt-5-0.746εt-4 for food poisoning persons were established and forecasting effect was optimized.True values were all within 95% CI of predicted ones.Conclusion ARIMA model can be well used to simulate incidence trend of food poisoning.