基于贝叶斯估计的空间Pair-copula预测模型及应用

Bayesian Estimation-Based Space Pair-copula Prediction Model and Its Application

为了描述空间相关关系的复杂性,文章结合Pair-copula函数将空间相关结构分解成双变量之间的条件相关。对Pair-copula空间分析模型提出了边缘分布参数及相关结构参数的贝叶斯估计,结合实际数据探讨了高斯随机场下的先验分布选择问题,给出了未知参数后验分布函数。提出了基于Pair-copula函数的空间预测方法,并通过交叉验证,与传统空间预测方法的预测精度进行了比较。结果表明,Pair-copula空间预测方法在对复杂空间数据的分析上具有明显的优势。

In order to describe the complexity of spatial correlations, this paper combines the pair-copula function to decompose the spatial correlation structure into conditional correlation between two variables. Then, the paper proposes Bayesian estimation of marginal distribution parameters and related structural parameters for pair-copula spatial analysis model, and uses actual data to discuss the problem of prior distribution selection under Gaussian random field, giving the posterior distribution function of unknown parameters. Finally, a spatial prediction method based on pair-copula function is proposed and compared with the prediction accuracy of traditional spatial prediction method by cross-validation. The results show that the pair-copula spatial prediction method has significant advantages in the analysis of complex spatial data.