scientific-adaptive-experiments

To optimize multi-armed bandit experiments, this Skill applies Thompson sampling and UCB logic to dynamically allocate experimental resources, speeding up convergence. It tracks outcomes continuously to balance exploration and exploitation.

Sequential analysis with SPRT evaluates accumulated observations incrementally to make statistical decisions. This Skill implements SPRT logic to enable early stopping, reducing unnecessary experimental trials once sufficient evidence is reached.

To run Bayesian dose finding and response-adaptive randomized trials, this Skill sequentially updates dose conditions based on observed data. It adapts treatment allocations dynamically during clinical trial design to optimize outcomes.

Yes, you can use this Skill for adaptive experimental planning without external dependencies. It independently implements Bayesian adaptive design, SPRT, and response-adaptive randomization rules to generate comprehensive trial summaries.

To compare Thompson sampling and UCB in adaptive experiments, this Skill automates both strategies within multi-armed bandit simulations. It tracks and generates comparative reports summarizing outcomes and convergence rates.

Bayesian adaptive design for sequential analysis may not suit experiments requiring rigid, fixed sample sizes set prior to observation. It dynamically updates conditions and stops early based on accumulating data, which conflicts with static protocols.