run-diagnostics

Model Execution and Diagnostics Workflow

You are helping the user run and diagnose their Bayesian model.

Step 1: Identify Model and Data

Determine:

1. Model language: Stan / JAGS / WinBUGS / PyMC
2. Data source:
   • User-provided data
   • Generate synthetic test data

Step 2: Generate Test Data (if needed)

Use @test-runner to create appropriate synthetic data:

For Regression Models

set.seed(42)
N <- 100
K <- 3
X <- matrix(rnorm(N * K), N, K)
true_beta <- c(0.5, -0.3, 0.8)
true_sigma <- 1
y <- X %*% true_beta + rnorm(N, 0, true_sigma)

stan_data <- list(N = N, K = K, X = X, y = as.vector(y))

For Hierarchical Models

set.seed(42)
J <- 8
N_per_group <- 20
true_mu <- 5
true_tau <- 3
true_theta <- rnorm(J, true_mu, true_tau)

y <- unlist(lapply(1:J, function(j) rnorm(N_per_group, true_theta[j], 2)))
group <- rep(1:J, each = N_per_group)

stan_data <- list(N = J * N_per_group, J = J, group = group, y = y)

Step 3: Execute Model

Stan (cmdstanr)

library(cmdstanr)

# Compile
mod <- cmdstan_model("model.stan")

# Short test run
fit <- mod$sample(
  data = stan_data,
  seed = 12345,
  chains = 2,
  parallel_chains = 2,
  iter_warmup = 500,
  iter_sampling = 500,
  refresh = 100
)

JAGS (R2jags)

library(R2jags)

fit <- jags(
  data = jags_data,
  parameters.to.save = c("mu", "sigma", "theta"),
  model.file = "model.txt",
  n.chains = 2,
  n.iter = 2000,
  n.burnin = 1000,
  progress.bar = "text"
)

PyMC (Python)

import pymc as pm
import arviz as az
import numpy as np

# Define model
with pm.Model() as model:
    # Priors
    mu = pm.Normal("mu", mu=0, sigma=10)
    sigma = pm.HalfNormal("sigma", sigma=1)

    # Likelihood
    y_obs = pm.Normal("y_obs", mu=mu, sigma=sigma, observed=y_data)

    # Sample
    trace = pm.sample(
        draws=1000,
        tune=1000,
        chains=2,
        cores=2,
        random_seed=12345,
        return_inferencedata=True
    )

# Diagnostics
print(az.summary(trace))
az.plot_trace(trace)