Two questions that sound the same but aren't | Archetype 1: one government's own trend | Archetype 2: a cross-government rollup | The mechanism | How big is the risk in practice | Why Total = Direct + M + L + Q, not Direct + M | Composition rules | Summary

Why per-year population matters | The four population sources | Coverage | The popyear quirk | Time-varying peer cohorts | Future direction

Introduction | Administrative Data Functions | Manipulating and Cleaning Data | Checking and Summarizing Data | Datasets | Deprecated functions

Introduction | Illustrating Model Effects | Random Effects | Substantive Effects | Uncertainty

Contextual effects | The data | A contextual-effects model | Predictions that separate the two effects | How much do schools matter across the rank distribution? | References

Introduction | Missing Data and its Discontents | Fitting and Summarizing a Model List | Output of a Model List | Specific Model Information Summaries | Diagnostics of List Components | Model List Generics | Cautions and Notes

Marginalizing Random Effects | Summarizing | Plotting

Introduction | Conceptual description | Comparison to existing methods | Step 1: Estimating the model and using predictInterval() | Step 1a: Adjusting for correlation between fixed and random effects | Step 2: Comparison with arm::sim() | Step 3: Comparison with lme4::bootMer() | Step 3a: lme4::bootMer() method 1 | Step 3b: lme4::bootMer() method 2 | Step 3c: lme4::bootMer() method 3 | Step 3d: Comparison to rstanarm | Computation time | Simulation

Why compare to brms? | A random-slopes model | Point estimates are essentially identical | The prediction intervals agree | And they are equally well calibrated | At a fraction of the cost | What about an entirely new group? | A generalized linear mixed model | Takeaways