Package: shrinkr 0.4.5

Introduction | Use Case: Multi-Hospital Mortality Prediction | Scenario Setup | The Federated Network | The Clinical Model | $$\text{logit}(\text{mortality}) | Federated Workflow | Step 1: Each Site Fits Independently | Step 2: Sites Share Summaries with Coordinator | Path A: Share Full Posteriors (if permitted) | Path B: Share Only Summary Statistics (requires assumptions) | Central Coordinator: Stage 2 Shrinkage | Path A: Using Full Posteriors | Specify Network-Level Priors | Fit Hierarchical Model | Path B: Using Only Summary Statistics | Step 1: Check Normality Assumption | When CLT Fails: A Counter-Example | Step 2: Fit Using CLT Approximation | Compare Paths | Results: Improved Site-Specific Estimates | Visualize Shrinkage Effect | Quantify Uncertainty Reduction | Visualize Uncertainty Reduction | Clinical Impact: Network-Calibrated Predictions | Stage 1: Independent Site Predictions | Stage 2: Network-Calibrated Predictions | Visualize Prediction Changes | Privacy-Preserving Benefits | What Gets Shared | Compliance Benefits | Advanced Federated Scenarios | Scenario 1: Heterogeneous Models | Scenario 2: Meta-Analysis of Published Studies | Scenario 3: Iterative Federated Updates | Federated Learning Best Practices | 1. Establish Data Governance | 2. Standardize Stage 1 Models | 3. Verify Normality if Using Summaries | 4. Quality Control | 5. Sensitivity Analysis | 6. Transparent Reporting | Advantages of shrinkr for Federated Learning | When to Use Federated shrinkr | Summary | Additional Resources | Session Info

Overview | The historical evidence | Hierarchical meta-analysis with shrinkr | Building the MAP prior | Robustify and form the posterior with beastt | Summary | References

What is shrinkr? | A Complete Example: Regional Clinical Trial | Stage 1: Fit Independent Models with Stan | Stage 2: Apply Hierarchical Shrinkage | Step 1: Fit Mixture Approximation | Step 2: Specify Hierarchical Priors | Step 3: Fit the Hierarchical Model | Step 4: Examine Results | Step 5: Visualize Shrinkage | Alternative Input: Using Summary Statistics Only | Complete Stan → shrinkr Workflow Summary | Key Concepts Checklist | Common Pitfalls to Avoid | Next Steps | Session Info

Overview | Setup | The Veteran Dataset | Approach 1: Two-Stage (brms + shrinkr) | Stage 1: Fit Cox Model | Stage 2: Apply Hierarchical Shrinkage | Step 1: Extract posterior samples | Step 2: Fit a Gaussian mixture approximation | Step 3: Apply a hierarchical prior | Approach 2: Full Hierarchical (brms) | Approach 3: Two-Stage (Frequentist + shrinkr) | Compare Three Approaches | Numerical comparison | Visual comparison | Sensitivity Analysis: Exploring Different Priors | Prior densities | Heterogeneity estimates | Impact on cell type estimates | Key Takeaways | Session Info

Overview | Example: Multi-Region Clinical Trial | Simulate Stage 1 Results | Fit shrinkr Model | Working with posterior Package | Extract Draws | Basic Summaries | Check Convergence | Diagnostic Plots with bayesplot | Trace Plots | Density Plots | Interval Plots | Area Plots | Tidy Analysis with tidybayes | Spread and Gather Draws | Point and Interval Summaries | Custom Summaries with dplyr | Computing Contrasts | Modern Visualizations with ggdist | Halfeye Plots | Slab + Interval | Quantile Dotplots | Gradient Intervals | Comparing Pre- and Post-Shrinkage | Extract Both Estimates | Custom Comparison Plot | Complete Workflow Example | Advanced: Custom Analyses | Probability Statements | Tail Probabilities | Ranking Analysis | Further Reading