各位好朋友：

不知道论坛里那位对贝叶斯统计软件winbugs比较熟悉，我是想写一篇meta-analysis的毕业论文，从文献“Minelli C, Thompson JR, Abrams KR, Lambert PC.Bayesian implementation of a genetic model-free approach to the meta-analysis of genetic association studies.Stat Med. 2005 Dec 30;24(24):3845-61”上看到一种方法，作者给出两个model，

Model1:
{

for( i in 1:n) {
      
case[i,1:3] ~ dmulti(p.case[i,], N.case)
control[i,1:3] ~ dmulti(p.ctrl[i,], N.ctrl)
sum <- p.ctrl[i,1] + exp(lambda*delta)*p.ctrl[i,2]+exp(delta)*p.ctrl[i,3]
p.case[i,1] <- p.ctrl[i,1]/sum
p.case[i,2] <- exp(lambda*delta)*p.ctrl[i,2]/sum
p.case[i,3] <- exp(delta)*p.ctrl[i,3]/sum
delta ~ dnorm(d, prec)
p.ctrl[i,1] <- exp(a)/(1 + exp(a)) * exp(b)/(1 + exp(b))
p.ctrl[i,2] <- exp(a)/(1 + exp(a)) - p.ctrl[i,1]
p.ctrl[i,3] <- 1- p.ctrl[i,1] - p.ctrl[i,2]
a ~ dnorm(0.0,1.0E-6)
b ~ dnorm(0.0,1.0E-6)                            
                     }
        
          d ~ dnorm(0.0,1.0E-6)
          lambda ~ dbeta(0.5,0.5)
                   prec <- 1/var
                   var <- pow(sd,2)
                   sd ~ dnorm(0,1)I(0,)
          OR_GG <- exp(d)
          OR_Gg <- exp(d*lambda)                                   
}
model2:
{
for( i in 1:n) {

   r_gg ~ dbin(p_gg, n_gg)
   r_Gg ~ dbin(p_Gg, n_Gg)
   r_GG ~ dbin(p_GG, n_GG)
   logit(p_gg) <- mu - delta[j]/2
   logit(p_GG) <- mu + delta[j]/2
   logit(p_Gg) <- mu + lambda*delta - delta[j]/2
   mu ~ dnorm(0.0,1.0E-5)
   delta ~ dnorm(d, prec)                
                     }
      
          lambda ~ dnorm(0.5,0.1)
          d ~ dnorm(0.0,1.0E-6)
          prec <- 1/(sd*sd)
          sd ~ dunif(0,100)
          OR_GG <- exp(d)
          OR_Gg <- exp(d*lambda)                                   
}
我的数据是：
Data (case group,gg:19,132,31,126,15,172,129,371,123,245; Gg:401,108,34,55,391,183,57,54,172,400; GG:44,167,153,145,129,402,132,143,106,131. control group, gg:157,99,128,154,138,179,70,170,156,156; Gg:59,150,63,157,140,90,141,413,91,128; GG:141,81,115,22,16,11,51,121,16,101)
主要思想就是OR1=(ggcase/GGcase)/(ggctrl/GGctrl), OR2=(Ggcase/GGcase)/(Ggctrl/GGctrl), lambda=logOR2/logOR1
上述两个model都是处理这三个指标的，不知道群里那位大大知道数据怎么输入，麻烦告诉我一下，本人做毕业论文急用，如能帮助解决，本人真是感激您一辈子！如果有什么地方没有说清楚，请和我联系，Email：chengducdc@163.com, QQ: 649161887.