Constantine
summary
Claims a Bayesian latent-variable model for mixed binary and continuous trial outcomes that combines effects into a single preference score using MCDA and propagates uncertainty; implements it with Stan code and notebook and models a latent multivariate Gaussian with observed links for binaries. Results give a modest advantage for treatment — about 62% probability a new patient scores higher — small posterior correlations and robustness to reasonable weight changes.
AI-written summary · about
chapters (6)
- 0:00Introduction
- 0:26Data Set
- 3:20General Framework
- 5:05Multicriteria decision analysis
- 12:39Correlation
- 14:10Questions
related talks
01h 06mNonparametric Bayesian Methods: Models, Algorithms, and Applications I
38mMixed and Covariate-Dependent Graphical Models
30mProbabilistic modelling with Scala
45mPreference Alignment from Choices and Response Times
01h 24mBayesianism and Survival Analysis
38mUsable A/B testing – A Bayesian approach
48mLearning Preferences with Latent State
48mAdaptivity and Confounding in Multi-armed Bandit Experiments