probability - Error:"Multiple definitions of node" in OpenBUGS. -
so thought following code work in openbugs, instead gives me "multiple definitions of node z" error.
model { z <- round(x) x ~ dnorm(0,1)t(-2,2) } list(z=0)
even if replace z <- round(x)
z <- x
continue same error. fact can deduce error resulting use of logical assignment observable variable , in particular, error not due round()
operation.
why bugs not allow this? also, work-around in case? here more general version want implement, modeling discrete gaussian walls (the truncation):
model { for(i in 1:n){ z[i] <- round(x[i]) x[i] ~ dnorm(mu,1)t(-2,2) } mu ~ dunif(-2,2) }
essentially, want z
distributed discrete gaussian "walls" (the truncation) , want estimate mu
data on z
. suppose can try make z
categorical variable , estimate parameters seems theoretically painful. there bugs trick can use intended model?
winbugs , openbugs don't allow observed data deterministic function of unobserved variable. suggest, use dcat() , express probabilities in terms of normal distribution.
though might prefer switch jags, has distribution dround() deals situation - data rounded n significant digits, in case n=0. though this forum post suggests there's bug in current stable release case, , might need download development version.
Comments
Post a Comment