Eve library

@phoebe/design

Other

Revision 1

On this page

Optimal experimental design and adaptive-trial machinery for Phoebe. Provides Bayesian Optimal Experimental Design (exact, Nested-Monte-Carlo, and PCE lower-bound Expected Information Gain), grid-Bayesian Adaptive Design Optimization (ADO) over psychometric, delay-discounting, and risk models, amortized greedy-BOED policies that serve designs in O(1), group-sequential efficacy boundaries (Lan-DeMets O'Brien-Fleming / Pocock spending and Haybittle-Peto), response-adaptive randomization (Thompson & RSIHR), BOIN dose-finding boundaries, predictive-probability futility, master-protocol (platform/basket/umbrella) operating-characteristic simulation, and SMART designs with backward-induction Q-learning value estimation of dynamic treatment regimes. All algorithms are deterministic under a seeded LCG PRNG — Date.now and Math.random are never used.