WinBUGS equals function with NA

Question

Would someone please let me know what the equals function in WinBUGS return if one of the elements in the equals brackets is NA?

Specifically, I want to do the transformation:

for(i in 1:N){
    newvar[i] <- 1*equals(oldvar[i],0) + 2*equals(oldvar[i],1.5) + 3*equals(oldvar[i],4) + 4*equals(oldvar[i],8)
}

and want to preserve NA's, so records with NA in oldvar[i] also has NA in newvar[i], but I am afraid that this tranformation turns NA's into 0's.

Any and all advice is much appreciated.

Trang

Answer 1

This should give an error message (made use of undefined node) if you haven't put a prior distribution on oldvar[1]. WinBUGS expects all nodes to have values if they are used in calculations. If you have put a prior on it, then WinBUGS should generate values from the posterior distribution of oldvar[1], and then the equals() function should operate on those generated values. Have you used a prior which systematically generates zeroes for oldvar[1]? If you're not interested in modelling the missing data, I'd advise dropping the missing values from your data before using it in WinBUGS.