CLIP-OGD: An Experimental Design for Adaptive Neyman Allocation in Sequential Experiments

Authors: Jessica Dai, Paula Gradu, Christopher Harshaw

NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental To complement our theoretical results, we conduct simulations using data from a microeconomic experiment. In Section 7, we support these theoretical results with simulations using data from a microeconomic experiment.
Researcher Affiliation Academia Jessica Dai UC Berkeley jessicadai@berkeley.edu Paula Gradu UC Berkeley gradu@berkeley.edu Christopher Harshaw MIT charshaw@mit.edu
Pseudocode Yes Algorithm 1: CLIP-OGD
Open Source Code Yes A repository for reproducing simulations is: https://github.com/crharshaw/Clip-OGD-sims
Open Datasets Yes We evaluate the performance of CLIP-OGD and Explore-then-Commit (ETC) for the purpose of Adaptive Neyman Allocation on the field experiment of Groh and Mc Kenzie [2016], which investigates the effect of macro-insurance on micro-enterprises in post-revolution Egypt.
Dataset Splits No The paper does not explicitly provide details about train, validation, or test dataset splits. It mentions using "the first T units in the sequence" as the population for a given T, and that "Units are shuffled to appear in an arbitrary order and outcomes are normalized."
Hardware Specification Yes Simulations were run on a 2019 Mac Book Pro with 2.4 GHz Quad-Core Intel Core i5 and 16 GB LPDDR3 RAM.
Software Dependencies No The paper does not provide specific software dependencies with version numbers (e.g., Python version, library versions) used for the experiments.
Experiment Setup Yes CLIP-OGD is run with the parameters recommended in Theorem 4.2 and ETC is run with T0 T 1{3 so that the exploration phase grows with T. Theorem 4.2 states parameter values η a 1{T and α a 5 logp Tq.