摘要
目的探讨应用ARIMA时间序列模型预测广西孕产妇死亡率的可行性,为广西继续降低孕产妇死亡率提供理论依据。方法基于2002—2006年广西的逐月孕产妇死亡率,采用非条件最小二乘法估计模型参数,按照残差不相关原则与简洁原则确定模型结构,依据AIC与SBC准则确定模型的拟合优度,建立预测广西孕产妇死亡率的最优ARIMA模型。用所得模型预测广西2007年的孕产妇死亡率,比较预测值与实际值的差异;再以2002—2007年的数据构建模型预测广西2008年的孕产妇死亡率。结果模型ARIMA(2,1,0)(0,1,1)12较好拟合了既往时间段孕产妇死亡率的时间序列,模型自回归参数(AR1=-0.708,AR2=-0.537)与季节滑动平均参数(SMA1=0.511)均有统计学意义(P<0.05),AIC=33.814,SBC=39.364,模型残差为白噪声(P>0.05),模型数学函数式为(1+0.708B)(1+0.537B2)(1-B)(1-B12)Zt=(1-0.511 B12)μt。2007年逐月孕产妇死亡率的预测值符合实际值的变动趋势,2007年孕产妇死亡率与实际值的相对误差率仅为2.08%。预测2008年广西的孕产妇死亡率为21.232/10万。结论ARIMA模型可以较好地拟合孕产妇死亡率的时间变化趋势,并用于预测未来的孕产妇死亡率,是一种短期预测精度较高的预测模型。
Objective To explore the feasibility for application of time series ARIMA model to predicting the maternal mortality ratio （MMR） in Guangxi Zhuang Autonomous Region so as to provide the theoretical basis for continuing to reduce the MMR. Methods ARIMA model was established based on the monthly MMR of Guangxi from 2002 to 2006. The parameters of the model were estimated through unconditional least squares method, the structure was determined according to criteria of residual un-correlation and concision, and the goodness-of-fit of model was determined through Akaike Information Criterion （AIC） and Schwarz Bayesian Criterion （BSC）. The constructed optimal model was applied to predict the MMR in 2007 of Guangxi and evaluate the validity of model through comparing the difference of predicted MMR and actual MMR. The MMR in 2008 was predicted by ARIMA model based on the MMR from 2002 to 2007. Results Model ARIMA（2,1,0） （0,1,1）,2 could preferably fit the time series of MMR with both autoregressive coefficient （AR1 = - 0. 708,AR2 = -0. 537） and seasonal moving average coefficient （SMA1 = 0. 511） being statistically significant （P〈0.05）. AIC and BIC were 33. 814 and 39. 364 respectively and predicting error was white noise （P〉 0.05）. The mathematic function was （1 ＋ 0. 708B） （1 ＋ 0. 537B2 ） （ 1 - B） （ 1 - B^12 ） Z, = （1 - 0.511 B^12 ）μ1. The predicted MMR in 2007 was consistent with the actual MMR with the relative error of 2. 08%. The predicted MMR in 2008 based on the MMR from 2002 to 2007 would be 21. 232 per 100 thousand. Conclusions ARIMA model can be used to fit the changes of MMR and to forecast the future MMR. It is a predicted model of high precision for short-time forecast.
出处
《复旦学报（医学版）》
CAS
CSCD
北大核心
2008年第6期799-805,共7页
Fudan University Journal of Medical Sciences
基金
欧盟第六框架计划资助项目(517746)
