r语言quantmond_R语言使用贝叶斯层次模型进行空间数据分析

Leroux

模型

该模型是使用矩阵的“混合”(Leroux等人的模型)

定义的，以定义潜在效应的精度矩阵：

\ [

\ Sigma ^ {-1} = [(1-\ lambda)I_n + \ lambda M]; \ \ lambda \

in(0,1)

\]

为了定义正确的模型，我们应采用矩阵\(C \)如下：

\ [

C = I_n – M; \ M = diag(n_i)– W

\]

然后，\(\ lambda_ {max} = 1 \)和

\ [

\ Sigma ^ {-1} =

\ frac {1} {\ tau}(I_n- \ frac {\ rho} {\ lambda_ {max}} C)=

\ frac {1} {\ tau}(I_n- \ rho(I_n – M))= \ frac {1} {\ tau}((1- \

rho)I_n + \ rho M)

\]

为了拟合模型，第一步是创建矩阵\(M \)：

我们可以检查最大特征值\(\ lambda_ {max} \)是一个：

max(eigen(Cmatrix)$values)

## [1] 1

## [1] 1

该模型与往常一样具有功能inla。注意，\(C \)矩阵使用参数

传递给f函数Cmatrix：

##

## Call:

## Time used:

## Pre = 0.236, Running = 0.695, Post = 0.0493, Total = 0.98

## Fixed effects:

## mean sd 0.025quant 0.5quant 0.975quant mode kld

## (Intercept) -0.128 0.448 -0.91 -0.128 0.656 -0.126 0.075

## AVGIDIST 0.325 0.122 0.08 0.327 0.561 0.330 0.000

##

## Random effects:

## Name Model

## ID Generic1 model

##

## Model hyperparameters:

## mean sd 0.025quant 0.5quant 0.975quant mode

## Precision for ID 2.720 1.098 1.27 2.489 5.480 2.106

## Beta for ID 0.865 0.142 0.47 0.915 0.997 0.996

##

## Expected number of effective parameters(stdev): 52.25(13.87)

## Number of equivalent replicates : 5.38

##

## Deviance Information Criterion (DIC) ...............: 903.14

## Deviance Information Criterion (DIC, saturated) ....: 373.77

## Effective number of parameters .....................: 53.12

##

## Watanabe-Akaike information criterion (WAIC) ...: 906.20

## Effective number of parameters .................: 48.19

##

## Marginal log-Likelihood: -474.94

## Posterior marginals for the linear predictor and

## the fitted values are computed