Researchers compare data assimilation and likelihood-based inference for agent-based models

A new paper systematically compares Data Assimilation (DA) and Likelihood-Based Inference (LBI) for estimating latent states in Agent-Based Models (ABMs). While DA is broadly applicable and good for aggregate predictions, LBI offers more precise agent-level inference by directly using the model's likelihood function. The study found LBI superior for individual-level forecasts, even with model mis-specification, whereas DA remains competitive for aggregate outcomes. AI

Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →

RANK_REASON Academic paper comparing two inference methods on a specific model type.

Read on arXiv cs.LG →

• Data Assimilation
• Likelihood-Based Inference
• Agent-Based Models
• Bounded-Confidence Model
• arXiv

• paper
• other