CODEWITHSUNDEEP
rlinear-regressionbayesianmcmcjags
jjw
jjw

Reputation: 498

jags bayesian linear regression, how can i set priors when the priors are dependent each other?

library(rjags)
jags_code = "model{
for (i in 1:n){
Y[i] ~ dnorm(mu[i], tau)
mu[i] <- beta0 + beta1*x[i]}

beta0 ~ dnorm(0,0.001)
beta1 ~ dnorm(0,0.001)
eta <- dgamma(0.1/2, 0.1/2)
tau ~ dgamma(5/2, 5*eta/2)"

jags_data = list(x = c(5,1,2,3,4),
                 Y = c(6,11,12,3,4), n=5)

jags_model = jags.model(textConnection(jags_code), data=jags_data)
update(jags_model, 10000)
samp <- coda.samples(jags_model, variable.names=c('beta0', 'beta1', 'eta'),
                     n.iter=5000)

summary(samp)
plot(samp)

Hi, I'm trying to run Gibbs sampler and fit linear regression using rjags. However, above code produces error like this.

Error in jags.model(textConnection(jags_code), data = jags_data) : 
Error parsing model file:
syntax error on line 10 near ""

I guess it's due to tau which contains eta in its argument. How can I solve this?

Upvotes: 0

Views: 266

Answers (1)

duckmayr
duckmayr

Reputation: 16920

The error message you got is because you didn't provide a closing curly brace to pair with the opening curly brace of the model statement. Once you fix that error, you'll also notice that you accidentally used eta <- rather than eta ~, which will cause the error

Incorrect number of arguments in function dgamma

So, the completely fixed code is

library(rjags)
jags_code = "model{
for (i in 1:n){
Y[i] ~ dnorm(mu[i], tau)
mu[i] <- beta0 + beta1*x[i]}

beta0 ~ dnorm(0,0.001)
beta1 ~ dnorm(0,0.001)
eta ~ dgamma(0.1/2, 0.1/2)
tau ~ dgamma(5/2, 5*eta/2)}"

jags_data = list(x = c(5,1,2,3,4),
                 Y = c(6,11,12,3,4), n=5)

jags_model = jags.model(textConnection(jags_code), data=jags_data)

which will run just fine

Upvotes: 2

Related Questions