I'm taking a maymester Bayes course right now and am having a hard time with running a multinomial logit model. My data is from 2005. Independents are: car type, urban residence, being married, being black, being female, age, and education. The dependent variable is a categorical variable measuring approval for President Bush (4 pt ordinal scale).

My model is syntactically correct according to WinBUGS (though I know that doesn't mean much). But when I try and load my data my error message is "Expected number or NA."

Can anyone see any glaring errors in my code? I'd welcome any suggestions.

Code:

```
model{
for(i in 1:N){
bushap[i] ~ dcat(p[i,1:J])
for(j in 1:J){
log(q[i,j]) <- b[1,j] + b[2,j]*Car[i] + b[3,j]*Educ[i] + b[4,j]*age[i] +
b[5,j]*urban[i] + b[6,j]*married[i] + b[7,j]*black[i] + b[8,j]*female[i]
p[i,j] <- q[ID,j]/sum(q[i,1:J])
pred[i,j] <- equals(p[i,j],ranked(p[i,1:4],4) )
}
predcat[i] <- pred[i,1] + pred[i,2]*2 + pred[i,3]*3 + pred[i,4]*4
}
for(k in 1:8){
b[k,1] <- 0
for(j in 2:J){
b[k,j] ~ dnorm(0, 0.1)
}
}
unused<- democ[1]
unused<- GOP[1]
unused<-kids[1]
}
```