Please use this identifier to cite or link to this item: https://hdl.handle.net/10419/337369 
Year of Publication: 
2026
Series/Report no.: 
GLO Discussion Paper No. 1717
Publisher: 
Global Labor Organization (GLO), Essen
Abstract: 
Researchers typically collect experimental data sequentially, allowing early outcome observations and adaptive treatment assignment to reduce exposure to inferior treatments. This article reviews multi-armed-bandit adaptive experimental designs that balance exploration and exploitation. Because adaptively collected experimental data through bandit algorithms violate standard asymptotics, inference is challenging. We implement an estimator that yields valid heteroskedasticity-robust confidence intervals in batched bandit designs and compare coverage in Monte Carlo simulations. We introduce bbandits for Stata, a tool for designing experiments via simulation, running interactive bandit experiments, and implementing and analyzing adaptively collected data. bbandits includes three common assignment algorithms-e-first, e-greedy, and Thompson sampling-and supports estimation, inference, and visualization.
Subjects: 
Randomized controlled trial
causal inference
multi-armed bandits
experimental design
machine learning
JEL: 
C1
C11
C12
C13
C15
C18
C8
C87
C88
C9
D83
Document Type: 
Working Paper

Files in This Item:
File
Size





Items in EconStor are protected by copyright, with all rights reserved, unless otherwise indicated.