节点与父母不一致，JAGS

时间：2017-02-19 23:29:56

标签： r jags

我试图估算JAGS中的模型。我希望州立大学能够依赖区域平均水平。我一直在搞乱它，除了错误之外别无其他任何东西。我无法理解为什么会发生这种情况，因为当我向法线的平均值添加一个常量时，它不会抛出任何错误（我在下面添加的1）但是当我添加另一个常量时，z [n]的我得到了节点错误。 

Error in node y[8]
Node inconsistent with parents

model{
for(i in 1:N)
{
  gend[i] = inprod(betaG, g[i])
  eta[i] = beta0  + betaState[state[i]] + gend[i]
  pi[i] = 1/(1 + exp(-eta[i]))

  y[i] ~ dbern(pi[i])
}

beta0 ~dnorm(0, .001)
betaG ~ dnorm(0, .001)

for( r in 1:4){
  region[r] ~ dnorm(0, .0001)
}

gamma <- c(1,1)

for(n in 1:9){
  z[n] <- inprod(gamma, xState[n, 2:3])
  betaState[n] ~ dnorm(region[1] + z[n], .001)
}

for(n in 10:26){
  betaState[n] ~ dnorm(region[2] +1, .001)
}

for(n in 27:38){
  betaState[n] ~ dnorm(region[3]+1 , .001)
}
for(n in 39:51){
  betaState[n] ~ dnorm(region[4]+ 1, .001)
}

}

我已经查看了有关此错误的其他人的帖子，我无法实现发布到我的问题的答案。

1 个答案:

答案 0 :(得分：0)

哇，我走错了路。我试图复制Shirley和Gelman关于死刑的论文，文中有模型。它们具有宏大的民族意味，影响着区域手段，影响着国家的手段。有很多交互效果，我不是100％关于这个，但我用jags编写分层模型的方式我会在这里发布，因为没有很多关于jags代码的在线示例：

model{
for(i in 1:N)
{
  asy[i] <- alphaStateYear[dataTime[i,2], state[i]]
  aes[i] <- alphaEdState[ed[i], rData[i], state[i]]
  aas[i] <- alphaAgeState[age[i], rData[i], state[i]]

  bbs[i] <- betaBlackState[race[i,3], rData[i], state[i]]
  bgs[i] <- betaGenderState[g[i,3], rData[i], state[i]]
  bbfs[i] <- betaBlackFemaleState[g[i,3], race[i,3], rData[i], state[i]]

  das[i] <- deltaAgeState[age[i], rData[i], state[i]] 
  dbs[i] <- deltaBlackState[race[i,3], rData[i], state[i]]
  dgs[i] <- deltaGenderState[g[i,3], rData[i], state[i]]
  dbfs[i] <- deltaBlackFemaleState[g[i,3], race[i,3], rData[i], state[i]]

  eta[i] <- (asy[i] +  aes[i] + aas[i])  +  
            (das[i] * dataTime[i, 1]) + 
            (bbs[i] * race[i, 2]) + (dbs[i] * race[i,2] * dataTime[i, 1]) +
            (bgs[i] * g[i,2]) + (dgs[i]*g[i,2]*dataTime[i,1]) + 
            (bbfs[i]*race[i,2]*dataTime[i,1]) + (dbfs[i] * race[i,2] * g[i,2] * dataTime[i,1])

  pi[i] = 1/(1 + exp(-eta[i]))

  y[i] ~ dbern(pi[i])
}


for( r in 1:4){
  alphaRegion[r] ~ dnorm(0, .0001)
  deltaRegion[r] ~ dnorm(0, .0001)
  sigiRegion[r] ~ dt(0,5,3)I(0,)
}

for( k in 1:51){
  mu[k] <- alphaRegion[region[k]] + inprod(beta[1:2],xState[k, 1:2])
  alphaState[k] ~ dnorm(mu[k], sigiRegion[region[k]])
}

beta[1] ~ dnorm(0, .04)
beta[2] ~ dnorm(0, .04)


nationalMu ~ dnorm(0, .04)
muDelta ~ dnorm(0, .04)
phi ~ dunif(-1,1)
sigma ~ dt(0, 5, 3)I(0,)
mu1[1] <- nationalMu + muDelta * timeAndIndex[1, 2]

alphaYear[1] ~ dnorm(mu1[1], pow( pow(sigma,2)/ (1- pow(phi,2)), -1))
for( t in 2:54 ){
  mu1[t] <- nationalMu + muDelta * timeAndIndex[t, 2] + phi*(alphaYear[t-1] - nationalMu - muDelta * timeAndIndex[t-1, 2])
  alphaYear[t] ~ dnorm(mu1[t], pow(sigma,-2))
}

sigmaStateYear ~ dt(0, 5, 3)I(0,)
for(yr in 1:54){
  for(st in 1:51){
    alphaStateYear[yr, st] ~ dnorm(mu2[yr, st], sigmaStateYear)
    mu2[yr, st] <- alphaYear[yr] + alphaState[st] + deltaState[st] * timeAndIndex[yr, 2]
  }
}


tauStateYear[1] ~ dt(0, 5, 3)I(0,)
tauStateYear[2] ~ dt(0, 5, 3)I(0,)
tauStateYear[3] ~ dt(0, 5, 3)I(0,)
tauStateYear[4] ~ dt(0, 5, 3)I(0,)
for(q in 1:51){
  deltaState[q] ~ dnorm(mu3[q], tauStateYear[region[q]])
  mu3[q] <- deltaRegion[region[q]] + inprod(gamma[1:2], zState[q, 1:2])
}

gamma[1] ~ dnorm(0, .04)
gamma[2] ~ dnorm(0, .04)


for( a in 1:4){
  nationalAge[a] ~ dnorm(0, .04)
  nationalAge1[a] ~ dnorm(0, .04)
  nationalTau[a] ~ dt(0, 5, 3)I(0,)
  nationalSigma[a] ~ dt(0, 5, 3)I(0,)
  for(indx in 1:4){
    regionalAge[a, indx] ~ dnorm(nationalAge[a], pow(nationalSigma[a],-2))
    regionalAge1[a, indx] ~ dnorm(nationalAge1[a], pow(nationalTau[a],-2))
    regionalTau[a, indx] ~ dt(0, 5, 3)I(0,)
    regionalSigma[a, indx] ~ dt(0, 5, 3)I(0,)
    for(indx2 in 1:51){
      alphaAgeState[a, indx, indx2] ~ dnorm(regionalAge[a, indx], pow(regionalSigma[a, indx],-2))
      deltaAgeState[a, indx, indx2] ~ dnorm(regionalAge[a, indx], pow(regionalTau[a, indx],-2))
    }
  }
}

for( e in 1:4){
  nationalEd[e] ~ dnorm(0, .04)
  nationalSigmaEd[e] ~ dt(0, 5, 3)I(0,)
  for(indxA in 1:4){
    regionalEd[e, indxA] ~ dnorm(nationalEd[e], pow(nationalSigmaEd[e],-2))
    regionalSigmaEd[e, indxA] ~ dt(0, 5, 3)I(0,)
    for(indxB in 1:51){
      alphaEdState[e, indxA, indxB] ~ dnorm(regionalEd[e, indxA], pow(regionalSigmaEd[e, indxA],-2))
    }
  }
}

for(blackNonBlack in 1:2){
  blackGrandMean[blackNonBlack] ~ dnorm(0, sig)
  deltaGrandMean[blackNonBlack] ~ dnorm(0, sig)
  nationalSigmaBlack[blackNonBlack] ~ dt(0, 5, 3)I(0,)
  nationalTauBlack[blackNonBlack] ~ dt(0, 5, 3)I(0,)
  for(reg in 1:4){
    blackReg[blackNonBlack, reg] ~ dnorm(blackGrandMean[blackNonBlack], nationalSigmaBlack[blackNonBlack])
    deltaReg[blackNonBlack, reg] ~ dnorm(deltaGrandMean[blackNonBlack], nationalTauBlack[blackNonBlack])
    regionalSigmaBlack[blackNonBlack, reg] ~ dt(0, 5, 3)I(0,)
    regionalTauBlack[blackNonBlack, reg] ~ dt(0, 5, 3)I(0,)
    for(sta in 1:51){
      betaBlackState[blackNonBlack, reg, sta] ~ dnorm(blackReg[blackNonBlack, reg], regionalSigmaBlack[blackNonBlack, reg])
      deltaBlackState[blackNonBlack, reg, sta] ~ dnorm(deltaReg[blackNonBlack, reg], regionalTauBlack[blackNonBlack, reg])
    }
  }
}

for(gender in 1:2){
  genderGrandMean[gender] ~ dnorm(0 , .04)
  deltaGenderGrandMean[gender] ~ dnorm(0, .04)
  nationalGenderSigma[gender] ~ dt(0, 5, 3)I(0,)
  nationalGenderTau[gender] ~ dt(0, 5, 3)I(0,)
  for(regi in 1:4){
    genderReg[gender, regi] ~ dnorm(genderGrandMean[gender], nationalGenderSigma[gender])
    deltaGenderReg[gender, regi] ~  dnorm(deltaGenderGrandMean[gender], sig)
    regionalGenderSigma[gender, regi] ~ dt(0, 5, 3)I(0,)
    regionalGenderTau[gender, regi] ~ dt(0, 5, 3)I(0,)
    for(place in 1:51){
      betaGenderState[gender, regi, place] ~ dnorm(genderReg[gender, regi], regionalGenderSigma[gender, regi])
      deltaGenderState[gender, regi, place] ~ dnorm(deltaGenderReg[gender, regi], regionalGenderTau[gender, regi])
    }
  }
}

sig <- .1
for (gen in 1:2){
  ggm[gen] ~ dnorm(0, sig)
  dggm[gen] ~ dnorm(0, sig)
  nationalGenSigma[gen] ~ dt(0, 5, 3)I(0,)
  nationalGenTau[gen] ~ dt(0, 5, 3)I(0,)
  for(ra in 1:2){
    imBlack[gen, ra] ~ dnorm(ggm[gen], nationalGenSigma[gen])
    deltaImBlack[gen, ra] ~ dnorm(dggm[gen], nationalGenTau[gen])
    genderRaceSigma[gen, ra] ~ dt(0, 5, 3)I(0,)
    genderRaceTau[gen, ra] ~ dt(0, 5, 3)I(0,)
    for(part in 1:4){
      imBlackInPart[gen, ra, part] ~ dnorm(imBlack[gen, ra], genderRaceSigma[gen, ra])
      deltaImBlackPart[gen, ra, part] ~ dnorm(deltaImBlack[gen, ra], genderRaceTau[gen, ra])
      genderRaceRegionSigma[gen, ra, part] ~ dt(0, 5, 3)I(0,)
      genderRaceRegionTau[gen, ra, part] ~ dt(0, 5, 3)I(0,)
      for(estado in 1:51){
        betaBlackFemaleState[gen, ra, part, estado] ~ dnorm(imBlackInPart[gen, ra, part], genderRaceRegionSigma[gen, ra, part])
        deltaBlackFemaleState[gen, ra, part, estado] ~ dnorm(deltaImBlackPart[gen, ra, part], genderRaceRegionTau[gen, ra, part])
      }
    }
  }
}

}
相关问题
最新问题