HSI
From: Bayesian Models for Astrophysical Data, Cambridge Univ. Press
(c) 2017, Joseph M. Hilbe, Rafael S. de Souza and Emille E. O. Ishida
you are kindly asked to include the complete citation if you used this material in a publication
​
​
Code 7.1 Bayesian zero-inflated Poisson model in R using JAGS
====================================================
require(MASS)
require(R2jags)
require(VGAM)
​
set.seed(141)
nobs <- 1000
​
x1 <- runif(nobs)
xb <- 1 + 2.0*x1
xc <- 2 - 5.0*x1
exb <- exp(xb)
exc <- 1/(1+exp(-xc))
zipy <- rzipois(n=nobs, lambda=exb, pstr0=exc)
zipdata <- data.frame(zipy,x1)
​
Xc <- model.matrix(~ 1 + x1, data=zipdata)
Xb <- model.matrix(~ 1 + x1, data=zipdata)
Kc <- ncol(Xc)
Kb <- ncol(Xb)
​
model.data <- list(Y = zipdata$zipy, # response
Xc = Xc, # covariates
Kc = Kc, # number of betas
Xb = Xb, # covariates
Kb = Kb, # number of gammas
N = nobs)
​
ZIPOIS<-"model{
# Priors - count and binary components
for (i in 1:Kc) { beta[i] ~ dnorm(0, 0.0001)}
for (i in 1:Kb) { gamma[i] ~ dnorm(0, 0.0001)}
​
# Likelihood
for (i in 1:N) {
W[i] ~ dbern(1 - Pi[i])
Y[i] ~ dpois(W[i] * mu[i])
log(mu[i]) <- inprod(beta[], Xc[i,])
logit(Pi[i]) <- inprod(gamma[], Xb[i,])
}
}"
​
W <- zipdata$zipy
W[zipdata$zipy > 0] <- 1
​
inits <- function() {
list(beta = rnorm(Kc, 0, 0.1),
gamma = rnorm(Kb, 0, 0.1),
W = W)}
​
params <- c("beta", "gamma")
​
ZIP <- jags(data = model.data,
inits = inits,
parameters = params,
model = textConnection(ZIPOIS),
n.thin = 1,
n.chains = 3,
n.burnin = 4000,
n.iter = 5000)
​
print(ZIP, intervals = c(0.025, 0.975), digits=3)
====================================================
Output on screen:
​
Inference for Bugs model at "3", fit using jags, 3 chains,
each with 5000 iterations (first 4000 discarded)
n.sims = 3000 iterations saved
mu.vect sd.vect 2.5% 97.5% Rhat n.eff
beta[1] 1.019 0.045 0.932 1.104 1.103 25
beta[2] 1.955 0.058 1.843 2.066 1.096 26
gamma[1] 1.913 0.166 1.588 2.241 1.013 160
gamma[2] -4.828 0.319 -5.472 -4.208 1.014 150
deviance 3095.598 11.563 3079.409 3122.313 1.004 710
For each parameter, n.eff is a crude measure of effective sample size,
and Rhat is the potential scale reduction factor (at convergence, Rhat=1).
DIC info (using the rule, pD = var(deviance)/2)
pD = 66.7 and DIC = 3162.3
DIC is an estimate of expected predictive error (lower deviance is better).