Learning Bayesian Statistics

#133 Making Models More Efficient & Flexible, with Sean Pinkney & Adrian Seyboldt

May 28, 2025
Guest
Adrian Seyboldt
Guest
Sean Pinkney
Sean Pinkney, a managing director at Omnicom Media Group and Stan contributor, teams up with Adrian Seyboldt, creator of NutBuy, to delve into innovative statistical modeling. They discuss enhancing hierarchical models with zero-sum constraints and the vital differences between population and sample means. Insights on Cholesky parameterization and improved sampling techniques are also explored. Their collaboration emphasizes how sharing knowledge fosters research advancements, making complex statistical problems more approachable and efficient.
Ask episode
AI Snips
Chapters
Transcript
Episode notes
INSIGHT

Zero-Sum Constraints Enhance Sampling

  • The zero-sum constraint in hierarchical models improves sampling efficiency.
  • It can be efficiently implemented with just one loop, avoiding costly matrix operations.
ADVICE

Predict New Groups Using Priors

  • Use population parameter estimates to predict for new groups in zero-sum hierarchical models.
  • Combining priors with these estimates can improve predictions for new group members.
INSIGHT

Population vs Sample Mean Effects

  • Distinguishing between population and sample means clarifies model parameter interpretations.
  • Zero-sum normal effects relate to deviations from sample means, enabling better hierarchical modeling.
Get the Snipd Podcast app to discover more snips from this episode
Get the app