CODEWITHSUNDEEP
rwinbugsmcmcr2winbugs
PepsiCo
PepsiCo

Reputation: 1409

error in using WinBUGS from R

The code is following:

require(BRugs)
require(R2WinBUGS)
model<-function(){
  for(i in 1:N){
  y[i] ~ dnorm(x[i], sigma.y)
  }
  x[1] ~ dnorm(theta[1], sigma.y)
  theta[1] <- 0
  for(j in 2:N){
    x[j] ~ dnorm(theta[j], sigma.x)
    theta[j] <- b*x[j-1] # this row wrong, 
# it would be right when I set theta[j] <- 1*x[j-1]
  }
  a ~ dunif(0, 1)
  b ~ dunif(-1, 1)
  sigma.y ~ dgamma(0.1, 0.1)
  sigma.x ~ dgamma(0.1, 0.1)
}
data <- list( N <- 100, y <- rnorm(100))
data=list(N=100,y=rnorm(100))

inits=function(){
  list(sigma.x = rgamma(1,0.1,0.1), sigma.y = rgamma(1, 0.1, 0.1), a = dnorm(1, 0, 1), b = dnorm(1, -1, 1))
}
parameters=c("a", "b", "x")

write.model(model, con = "model.bug")
modelCheck("model.bug")
# model is syntactically correct
ret.sim <- bugs(data, inits, parameters, "model.bug",
                n.chains = 1, n.iter = 1000,
                program= "winbugs",
                working.directory = NULL,
                debug = T)

I don't know why, the program will be correct when I replace theta[j] <- b*x[j-1] with theta[j] <- 1*x[j-1], but I have defined b ~ dunif(-1, 1). Indeed, I need to set theta[j] <- a - b*x[j-1] in the final model, and it turns out to be wrong when I try to add a and b into it. Anyone find where the problem is ?

Upvotes: 2

Views: 324

Answers (1)

guyabel
guyabel

Reputation: 8366

The problems is in your priors for b (and most likely a). I don't know your data but perhaps the range of your current priors do not include true values of a and b. I would think that if you use a continuous distribution(s):

a ~ dnorm(0,1)
b ~ dnorm(0,1)

your problem might be solved?

n.b. If you are trying to create a AR(1) model for WinBUGS you might want to check out the tsbugs package.

Upvotes: 2

Related Questions