pymc-bayesian-modeling
Build Bayesian Models with PyMC
Also available from: davila7
Build, fit, and validate Bayesian models using PyMC. Create hierarchical models, run MCMC sampling, and compare models with LOO and WAIC metrics.
Download the skill ZIP
Upload in Claude
Go to Settings → Capabilities → Skills → Upload skill
Toggle on and start using
Test it
Using "pymc-bayesian-modeling". Build a Bayesian linear regression with predictors X1, X2 and outcome y
Expected outcome:
- Model: alpha ~ Normal(0,1), beta ~ Normal(0,1), sigma ~ HalfNormal(1)
- Posterior means with 95% credible intervals
- R-hat values all below 1.01 indicating convergence
- Posterior predictive checks showing model fit
Using "pymc-bayesian-modeling". Create a hierarchical model with 10 groups and 50 observations each
Expected outcome:
- Population-level hyperparameters: mu_alpha, sigma_alpha, mu_beta, sigma_beta
- Group-level parameters showing shrinkage toward population mean
- Trace plots confirming good mixing
- Comparison of group intercepts with forest plot
Using "pymc-bayesian-modeling". Compare three model specifications using LOO
Expected outcome:
- LOO comparison table with rank, elpd_loo, and weights
- Pareto-k diagnostics for each model
- Recommendation for best model with interpretation
- Model-averaged predictions if models are similar
Security Audit
SafeAll 383 static findings are false positives. The 'weak cryptographic algorithm' detections flag legitimate PyMC probability distributions. The 'external_commands' findings flag markdown backtick syntax. This is a legitimate scientific computing skill for Bayesian statistical modeling.
Risk Factors
⚙️ External commands (4)
📁 Filesystem access (2)
Quality Score
What You Can Build
Uncertainty Quantification
Quantify uncertainty in regression coefficients and predictions using full posterior distributions.
Hierarchical Data Analysis
Analyze grouped or nested data structures with partial pooling across groups.
Model Comparison
Compare competing statistical models using LOO cross-validation and model weights.
Try These Prompts
Build a Bayesian linear regression model using PyMC with the following predictors and outcome variable. Include prior predictive checks, sample the posterior using NUTS, and check convergence diagnostics.
Create a hierarchical Bayesian model for data with group structure. Use non-centered parameterization to avoid sampling issues. Show group-level estimates and shrinkage toward the population mean.
Fit two or more Bayesian models to the same data and compare them using LOO. Report the model weights and provide interpretation guidelines for the comparison results.
Run a complete Bayesian analysis workflow including data preparation, model building, prior predictive checks, MCMC sampling with 4 chains, convergence diagnostics, posterior predictive checks, and predictions for new data.
Best Practices
- Standardize predictors before modeling to improve sampling efficiency
- Use weakly informative priors rather than flat priors for better inference
- Always run prior predictive checks before fitting to validate priors
- Check convergence diagnostics (R-hat < 1.01, ESS > 400, no divergences) before interpreting results
Avoid
- Using flat priors (Uniform(0, 1e10)) instead of informative priors
- Interpreting results without checking convergence diagnostics
- Using centered parameterization for hierarchical models (causes divergences)
- Fitting complex models without prior predictive checks