JimSavage
summary
Implements an aggregate random coefficients logit in Stan to handle only market-level shares while modeling individual heterogeneity with multivariate normal betas and Gumbel epsilons; addresses price endogeneity by embedding a simple structural supply-side markups rule (profit‑maximizing first‑order condition) rather than relying on external instruments; uses a fake-people Monte Carlo matrix draw to integrate over tastes. Notebook recovers most parameters from simulated data, with a noted bias in the effective price estimate, and argues the approach is identifiable.
AI-written summary · about
chapters (5)
- 0:00Start
- 3:56aggregate data only
- 5:34endogeneity in price-setting
- 6:39strategy in price setting
- 7:53The aggregate random coefficients logit model
related talks
47mIncrementality Bidding and Attribution
44mComplementary Information and Learning Traps
17mMaking sense of marginal effects
25mScaling Bayesian Experimentation
43mToward Human-Centered Machine Learning
26mThe specifics of the AB-test for Unit Economics with a network effect
45mThe Statistical Fairness-Accuracy Frontier
01h 17mLightning Talks