Linzer 2013 - Dynamic Bayesian Forecasting of Presidential Elections
Routing Summary
Full capture of Linzer (2013), JASA 108(501): 124-134, plus one concept note on extensions (E40, E94, E95). 5 notes.
- Need the paper on a page and section map? → Linzer 2013 - Dynamic Bayesian Forecasting of Presidential Elections - Overview
- Need the binomial / logit / reverse random-walk model and priors? → Linzer 2013 - Reverse Random-Walk Model and Priors
- Need state win probability or Electoral College simulation? → Linzer 2013 - Forecast Computation and Electoral College Simulation
- Need the 2008 data, priors, MAD, calibration? → Linzer 2013 - 2008 Application and Results
- Need house effects, correlated states, election-night reweighting? → Linzer 2013 - Extensions - House Effects, Correlated Errors and Election-Night Updating (Concept Note)
Concept Map
| Concept | Note | Type | Depends On | Key Result |
|---|---|---|---|---|
| Reverse random walk | Linzer 2013 - Reverse Random-Walk Model and Priors | definition | Linear-Gaussian State-Space Models | polls to fundamentals as early |
| EC probability | Linzer 2013 - Forecast Computation and Electoral College Simulation | definition | posterior draws | from joint draws |
| 2008 results | Linzer 2013 - 2008 Application and Results | example | model | MAD 1.4%, 338-387 EV |
| Importance reweighting | Linzer 2013 - Extensions - House Effects, Correlated Errors and Election-Night Updating (Concept Note) | theorem | frozen draws | weights = interval likelihoods, ESS |
Notes
- Linzer 2013 - Dynamic Bayesian Forecasting of Presidential Elections - Overview — CONTAINS: summary, E40/E94/E95 links, Source Map.
- Linzer 2013 - Reverse Random-Walk Model and Priors — CONTAINS: Eqs 1-3, windowing, PyMC sketch.
- Linzer 2013 - Forecast Computation and Electoral College Simulation — CONTAINS: state and EC probabilities, swing states, Python.
- Linzer 2013 - 2008 Application and Results — CONTAINS: data, priors table, results, calibration, trends.
- Linzer 2013 - Extensions - House Effects, Correlated Errors and Election-Night Updating (Concept Note) — CONTAINS: house effects, correlated states, learned prior, reweighting theorem, Python.
Sources
- Linzer 2013 - Dynamic Bayesian Forecasting of Presidential Elections in the States