【摘要】 本文建立了23个综合考虑水稻单产的时间效应、空间效应和时空交互效应的时空模型。利用湖北省1991-2007年县级水稻单产数据,借助WinBUGS进行Gibbs抽样估计,根据DIC准则进行选择最优模型,预测1992-2009年各县市的水稻单产。基于2008年预测单产及其后验分布,厘定县级水稻产量保险的纯费率。研究的主要结论:1992-2007年的预测值与实际值比较接近,且预测的水稻单产的标准差都比较小,表明所选择模型具有良好的短期预测能力;相邻地域的纯费率比较接近;所厘定的纯费率与蒙特卡洛(Monte Carlo,MC)误差正相关,Pearson相关系数为0.6793,MC误差包含来自时间、空间以及两者相互作用带来的不确定性。更多还原

【Abstract】 The article builds temporal-spatial models with consideration of the temporal trend,spatial effect and spatial-temporal effect.Using Hubei province county-level rice yields from 1991 to 2007,the article selects the best one from 23 models according to the DIC standard by Gibbs Sampling with WinBUGS and predicts rice yield of each county in 1992 and 2009.Based on the predicted yield and its posterior distribution in 2008,it calculates the premium rate of rice yield insurance at county level.The main conclusions are:the predicted yields of 1992-2007 are close to the actual yields,and the standard errors are small,which show that the selected model has a good capacity of prediction in short term. The premiums among neighboring counties are similar.The Pearson correlation coefficient between the pure premium rates and Monte Carlo errors is 0.6793,and the MC error contains the yields’ uncertainty from time,space,and spatial-temporal effect.更多还原