The stories behind our published research from last year

2026-01-06 · Source: Statistical Modeling, Causal Inference, and Social Science · Field: Science & Research — Mathematics & Computational Sciences, Artificial Intelligence & Machine Learning, Research Methodology & Innovation · Depth: Expert, long

Summary

Andrew Gelman has published a retrospective of his research from the past year, detailing 22 papers published in 2025 and 2026 across various statistical and social science journals. These publications cover diverse topics including adaptive sequential Monte Carlo for cross-validation, reanalysis of economic innovation models, statistical graphics, Bayesian workflow, and adjusting for underreporting in social studies. Other works address climate-fueled migration, AI and aesthetic judgment, meta-analysis with single studies, and the concept of normative scientific conflict. The collection also includes papers on stochastic potential outcomes, multilevel regression for HIV outcomes, statistical graphics and comics, simulation-based calibration for Bayesian computation, visualizing covariance matrices, and hierarchical Bayesian models for mitigating prediction disparities. Additionally, Gelman highlights several unpublished works and upcoming books, including "Bayesian Workflow" and a second edition of the "Handbook of Markov Chain Monte Carlo."

Key takeaway

For AI Scientists and statisticians grappling with complex modeling challenges, reviewing Gelman's recent publications offers practical insights into advanced Bayesian techniques and statistical workflow. You should consider how simulation-based calibration can validate your Bayesian computations and explore hierarchical Bayesian models to address prediction disparities. The emphasis on reanalysis and critical examination of scientific claims also provides a valuable framework for strengthening your own research robustness.

Key insights

Gelman's diverse research portfolio emphasizes Bayesian methods, statistical workflow, and critical re-evaluation of scientific practices.

Principles

• Bayesian workflow extends beyond Bayesian statistics.
• Normative scientific conflict is unavoidable and valuable.
• Generative models clarify complex statistical problems.

Method

The author's research often involves applying multilevel Bayesian approaches, reanalyzing existing studies, developing new statistical graphics techniques, and employing simulation-based calibration for computational checks.

In practice

• Explore adaptive sequential Monte Carlo for complex models.
• Consider "ladder of abstraction" for statistical graphics.
• Apply multilevel regression and poststratification for population inference.

Topics

• Bayesian Methods
• Statistical Graphics
• Causal Inference
• AI/ML Applications
• Statistical Workflow

Best for: AI Scientist, AI Researcher, Data Scientist, Research Scientist

Related on AIssential

• The Bayesian Workflow book is coming! — The "Bayesian Workflow" book offers practical guidance for applying Bayesian methods, addressing common challenges in statistical and computational practice.
• Forecasting Scientific Progress with Artificial Intelligence — Current AI systems cannot reliably forecast scientific progress or its timing, despite identifying plausible research directions.
• Complex & Intelligent Systems, Volume 12, Issue 2, February 2026 — The journal issue highlights diverse AI applications, from generative models to robust detection and optimization in complex systems.
• Some Open Problems in Probability that are Relevant to Applied Statistics (my talk this Wed noon at the Columbia statistics department student seminar) — Open problems in probability theory are crucial for advancing statistical practice and improving real-world predictions.
• UnpredictaBench: A Benchmark for Evaluating Distributional Randomness in LLMs — LLMs struggle significantly with generating samples calibrated to true underlying distributions, hindering their use in simulations.
• AutoGrad Changed Everything (Not Transformers) [Dr. Jeff Beck] — AutoGrad, not Transformers, fundamentally changed AI development by making it an engineering problem, enabling rapid experimentation.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Statistical Modeling, Causal Inference, and Social Science.