Monotone likelihood is a common issue in political science duration modeling, often leading to divergent parameter estimates. This problem arises when covariate values change monotonically with respect to failure time. Using mathematical explanations and simulations, this article demonstrates how Firth's penalized maximum likelihood estimation provides solutions.
Data & Methods Used
• Mathematical exposition
• Monte Carlo simulations
• Empirical political science applications
Key Findings
• Conditions for acute monotone likelihood identified
• Guidance for applying duration modeling techniques provided to avoid divergence toward infinity
Why It Matters
This research helps prevent misinterpretation and model misspecification in analyzing political events.
| Addressing Monotone Likelihood in Duration Modeling of Political Events was authored by Noel Anderson, Benjamin Bagozzi and Ore Koren. It was published by Cambridge in BJPS in 2021. |
