Inverse-Variance Weighting for Estimation of Heterogeneous Treatment Effects

Authors: Aaron Fisher

ICML 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We demonstrate the superior performance of IVWs in simulations
Researcher Affiliation Collaboration 1Genentech, Boston, MA, United States. Correspondence to: Aaron Fisher <afishe27@alumni.jh.edu>.
Pseudocode No The paper does not contain any clearly labeled pseudocode or algorithm blocks. It describes procedures through mathematical equations and textual explanations.
Open Source Code Yes Code for reproducing the methods and simulations in this paper is available at https://github.com/aaronjfisher/wpor/.
Open Datasets Yes We include 6 simulation scenarios, labeled A, B, C, D, E & F. The first four are experiments taken from Nie & Wager (2020), with |X| set equal to 10. Setting E is the low dimensional simulated example from Kennedy (2023). Setting F is the simple illustrative example from Figure 1. ... Table 2 presents each setting in detail, and Table 3 gives a qualitative summary of each setting.
Dataset Splits Yes For example, for f U,ˆθ, we used 90% of the data to estimate the nuisance functions ˆθ, evaluated and stored f U,ˆθ(Zi) for the remaining 10%, and then repeated this process 10 times with different fold assignments to obtain a pseudo-outcome for every individual.
Hardware Specification No The paper does not provide specific details regarding the hardware used for running the experiments or simulations, such as GPU/CPU models or memory specifications.
Software Dependencies No The paper mentions using 'the lightgbm R package written by Shi et al. (2023)' but does not provide specific numerical version numbers for this or any other software dependencies.
Experiment Setup Yes We implemented POR with three pseudo-outcome functions: f U,ˆθ, f DR,ˆθ, and fcov,ˆθ. In each case we used 10-fold crossfitting. ... We used boosted trees to perform all of our nuisance regressions, as well as the final regression predicting pseudo-outcomes as a function of X. ... For each pseudo-outcome function, we considered a weighted and unweighted version.