Question

我遇到winbugs代码问题。我最近开始学习它，现在我想用mcmc方法编写一个用于预测某些变量的代码。等式是： R = a1 * U + a2 * B + a3 * D ^ a4 + a5 * S ^ a6 + a7。我有R，U，D，B，S和a1到a7的所有值是随机变量，我想估计它们。这是我写的代码，我知道这不是真的。任何人都可以帮我纠正吗？

    model 
    {
        for(i in 1 : N) {

                R [i] <- a1 * U [i] + a2 * B [i] + a3 * pow(D [i] , a4) + a5 * pow( S [i] , a6) + a7
                            }



    # priors:

        a1 ~ dnorm(0.0,1.0E-4)                 
        a2 ~ dnorm(0.0,1.0E-4)            
        a3 ~ dnorm(0.0,1.0E-4);           
        a4 ~ dnorm(0.0,1.0E-4)            
        a5 ~ dnorm(0.0,1.0E-4)            
        a6 ~ dnorm(0.0,1.0E-4)
        a7 ~ dnorm(0.0,1.0E-4)           

        }


list(N=120, R = c( 2.19, 2.12, 1.88, 2.81, 2.2, 2.37, 2.34, 2.9, 3.04, 3.07, 3.04, 2.95, 2.66, 2.3, 2.17, 2.87, 2.48, 2.02, 1.87, 2, 2.45, 2.18, 2.17, 2.09, 2.1, 2.05, 2.6, 2.46, 2.2, 2.14, 2.78, 2.03, 2.05, 2.47, 2.05, 2.47, 1.83, 2.08, 1.87, 2.1, 2.31, 2.43, 1.79, 2.17, 2.17, 1.65, 1.75, 1.87, 1.78, 1.78, 2.39, 1.28, 1.85, 2.2, 2.43, 2.13, 1.71, 2.04, 2.39, 2.14, 1.88, 1.88, 1.27, 2.14, 1.77, 2.14, 1.6, 2.05, 2.44, 2.42, 1.67, 2.12, 2.17, 2.28, 2.12, 2.47, 2, 1.5, 2.47, 2.35, 2.07, 2.28, 2.16, 2.61, 1.91, 2.26, 1.85, 1.58, 1.51, 2.36, 1.82, 1.5, 2.05, 1.84, 1.38, 1.96, 1.85, 1.95, 1.6, 1.88, 1.88, 2.11, 1.64, 1.51, 2.17, 2.16, 1.98, 2.09, 1.77, 2, 1.84, 2.15, 1.7, 1.46, 2.23, 1.87, 1.82, 2.64, 1.94, 1.97
),

       U =   c( 199.7, 199.3, 199, 189.7, 189, 168.3, 174.1, 177.9, 180.7, 184.1, 192.4, 191, 191, 194.5, 195.2, 188.3, 188.4, 193.1, 182.8, 182.4, 182.4, 192.7, 192.6, 164.1, 160.7, 160.7, 144.8, 140, 137.2, 136.6, 135.2, 136.2, 137, 137.4, 137, 137.4, 139.3, 139.4, 141, 140.7, 139.7, 138.9, 136.5, 134.2, 131.4, 130, 128.6, 130, 131, 131, 129, 130, 130.1, 130.2, 138.6, 138.8, 138.8, 138.7, 138.6, 137.4, 137.2, 136.2, 135.2, 134.8, 134.1, 133.3, 133.4, 134.1, 134.3, 134.3, 134.3, 133.9, 133.8, 143.4, 145.4, 146.2, 145.9, 137.5, 140.7, 145.5, 158.6, 159, 159.3, 159.3, 160.3, 159.6, 153.8, 154.5, 157.9, 160.3, 163.2, 164.3, 166.4, 173.1, 173.1, 174.1, 174.5, 175.2, 176, 176.8, 176.8, 177, 175.9, 175.2, 173.1, 156.9, 150.7, 147.6, 147, 145.6, 144.8, 140.7, 125.5, 123.1, 122.3, 121.7, 121.7, 121.7, 120.7, 119.7
),


       B=  c( 55, 55, 55, 56, 56, 58, 58, 58, 57, 57, 54, 54, 53, 52, 52, 46, 45, 43, 39, 39, 39, 42, 42, 46, 46, 46, 42, 43, 42, 41, 41, 40, 39, 39, 39, 39, 38, 38, 38, 38, 37, 37, 35, 34, 34, 34, 32, 32, 32, 32, 31, 31, 31, 31, 31, 31, 30, 30, 30, 30, 30, 30, 30, 30, 30, 30, 30, 32, 32, 32, 33, 33, 33, 33, 34, 35, 35, 36, 36, 38, 37, 36, 36, 36, 35, 36, 35, 33, 33, 32, 32, 32, 30, 31, 31, 31, 31, 30, 30, 30, 30, 30, 30, 30, 30, 30, 30, 30, 29, 29, 29, 28, 29, 29, 29, 29, 30, 30, 29, 29
),  
        D  =  c( 0.8, 1.6, 2, 0.2, 2, 1.6, 2, 0.4, 0.2, 0.4, 0.4, 0.4, 0.1, 0.4, 0.4, 0.2, 0.2, 0.8, 0.8, 0.8, 0.4, 0.4, 0.2, 0.8, 1.6, 1.6, 0.2, 0.1, 1.6, 0.8, 0.4, 0.4, 1.6, 0.8, 1.6, 0.8, 2, 2, 1.6, 1.6, 1.6, 0.8, 1.6, 2, 2, 1.6, 1.6, 1.6, 0.4, 0.4, 0.8, 0.8, 0.8, 0.8, 0.4, 0.8, 0.4, 0.8, 0.2, 1.6, 0.8, 0.4, 1.6, 0.4, 1.6, 0.8, 1.6, 0.8, 0.4, 0.4, 1.6, 0.4, 0.4, 1.6, 1.6, 0.4, 0.8, 1.6, 0.1, 0.8, 0.1, 0.2, 0.8, 0.8, 0.8, 0.4, 0.8, 2, 1.6, 0.1, 0.8, 0.4, 0.4, 0.8, 0.8, 0.8, 0.1, 0.4, 2, 1.6, 1.6, 0.4, 2, 2, 1.6, 0.8, 1.6, 2, 2, 1.6, 2, 1.6, 2, 0.8, 0.8, 0.8, 0.8, 0.8, 0.8, 0.2
), 

        S  =  c( 25, 21, 20, 42, 40, 41, 35, 61, 55, 49, 32, 34, 23, 33, 30, 29, 24, 41, 20, 66, 55, 55, 59, 19, 66, 31, 67, 46, 70, 62, 29, 32, 21, 46, 21, 46, 41, 40, 10, 28, 36, 47, 14, 41, 44, 15, 10, 15, 2, 11, 39, 2, 26, 32, 58, 55, 8, 17, 50, 69, 67, 45, 7, 62, 76, 47, 18, 19, 30, 40, 4, 81, 60, 33, 71, 32, 61, 10, 15, 52, 36, 74, 35, 78, 26, 20, 52, 52, 10, 70, 21, 4, 74, 33, 11, 37, 4, 56, 71, 88, 77, 89, 66, 79, 63, 51, 81, 44, 86, 72, 70, 68, 70, 10, 57, 36, 56, 80, 68, 60
))


list(a7 = 2, a1= 0, a2 = 0, a3 = 0, 
    a4 = 0, a5= 0, a6 = 0)

a1, a2, a3, a4, a5, a6, a7

Answer 1

这看起来像你正在进行回归：

$R_i \sim N(\mu_i, \sigma)$

$\mu_i$ 是a1，a2等的函数。但是，您的模型暗示：

$R_i = \mu_i$

最大的区别在于您的模型中没有错误或不确定性。你可以这样做：

model{

    for(i in 1:N) {
        mu[i] <- a1 * U [i] + a2 * B [i] + a3 * pow(D [i] , a4) + a5 * pow( S [i] , a6) + a7
        R [i] ~ dnorm(mu[i], prec)
    }

    # priors:
    prec ~ dgamma(.001, .001)
    a1 ~ dnorm(0.0,1.0E-4)                 
    a2 ~ dnorm(0.0,1.0E-4)            
    a3 ~ dnorm(0.0,1.0E-4)           
    a4 ~ dnorm(0.0,1.0E-4)            
    a5 ~ dnorm(0.0,1.0E-4)            
    a6 ~ dnorm(0.0,1.0E-4)
    a7 ~ dnorm(0.0,1.0E-4)           

}

请注意，这也引入了一个新参数，即分布的标准偏差，在此建模为精度。

winbugs中的MCMC代码

1 个答案: