文档介绍：本文是通过对area,perime ter,campac tn ess几个变量的贝叶斯建模，来查看他们对gro ovelength这个变量的影响.
并且对比rjagsR2jags和内置贝叶斯预测函数的结果。
读取数据
read dat
library("arm") summary(bayesglm(groovelength~area+perimeter+campactness,data=seed))
bayesglm(formula = groovelength ~ area + =seed)
perimeter + campactness,
data
Devianee
Min
-
Residuals:
1Q Median 3Q
- -
Max
Coefficients:
Estimate
- -
(Intercept) area perimeter campactness
Std. Error t value
-
-
Pr(>|t|)
-13
-08
< 2e-16
***
***
**
***
Signif. codes: 0 '***'
'**'
'*' '.' ' ' 1
(Dispersion parameter for gaussian
family taken to be )
degrees of freedom degrees of freedom
Null devianee: on 208 Residual devianee: on 205 AIC: -
Number of Fisher Scoring iterations: 6
从内置贝叶斯模型的结果来看，个变量同样是非常显著，因此模型的结果和回归模型 类似。然后我们使用BUGS/JAGS软件包来建立贝叶斯模型
使用BUGS/JAGS软件包来建立贝叶斯模型
library(R2jags)
library(coda)
library(emdbook) for () and ()
library(arm) for coefplot
library(lattice)
library(dotwhisker)
首先设置因变量
groovelength=seed$groovelength
N <- length(groovelength)
建立贝叶斯模型
jagsl <- jags(='',
parameters=c("area","perimeter","campactness","int"), data = list =seed$area, 'b' = seed$perimeter, 'c'= se