Using priorsense with JAGS • priorsense

library(R2jags)
library(posterior)
library(priorsense)

set.seed(123)

priorsense is compatible with models fit with either jagsUI or R2jags.

Consider the univariate normal model with unknown mu and sigma available viaexample_powerscale_model("univariate_normal"). In JAGS, lprior and log_lik variables can be defined as below. By also defining separate lprior_mu and lprior_sigma variables, it will be possible to check the sensitivity for each prior separately.

model <- example_powerscale_model("univariate_normal", language = "jags")

model {
  for(n in 1:N) {
    y[n] ~ dnorm(mu, tau)
    log_lik[n] <- logdensity.norm(y[n], mu, tau)
  }
  mu ~ dnorm(0, 1)
  sigma ~ dnorm(0, 1 / 2.5^2) T(0,)
  tau <- 1 / sigma^2

  lprior_mu <- logdensity.norm(mu, 0, 1)
  lprior_sigma <- logdensity.norm(sigma, 0, 1 / 2.5^2)
  lprior <- lprior_mu + lprior_sigma
}

R2jags::jags() or jagsUI::jags() can be used to fit the model. Ensure that the required variables are monitored.

model_con <- textConnection(model$model_code)
data <- model$data

# monitor parameters of interest along with log-likelihood and log-prior
variables <- c("mu", "sigma", "log_lik", "lprior", "lprior_mu", "lprior_sigma")

fit <- R2jags::jags(
  data = data,
  model.file = model_con,
  parameters.to.save = variables,
  n.chains = 4,
  DIC = FALSE,
  quiet = TRUE,
  progress.bar = "none"
)

Then the priorsense functions will work as usual.

powerscale_sensitivity(fit)

Sensitivity based on cjs_dist
Prior selection: all priors
Likelihood selection: all data

 variable prior likelihood                           diagnosis
       mu 0.376      0.524 potential prior-likelihood conflict
    sigma 0.252      0.468 potential prior-likelihood conflict

powerscale_sensitivity(fit, prior_selection = "sigma")

Sensitivity based on cjs_dist
Prior selection: sigma
Likelihood selection: all data

 variable prior likelihood diagnosis
       mu 0.003      0.524         -
    sigma 0.006      0.468         -

powerscale_sensitivity(fit, prior_selection = "mu")

Sensitivity based on cjs_dist
Prior selection: mu
Likelihood selection: all data

 variable prior likelihood                           diagnosis
       mu 0.379      0.524 potential prior-likelihood conflict
    sigma 0.258      0.468 potential prior-likelihood conflict

powerscale_plot_dens(fit)