Stable Estimation of Heterogeneous Treatment Effects
Authors: Anpeng Wu, Kun Kuang, Ruoxuan Xiong, Bo Li, Fei Wu
ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Empirical results on both synthetic and real-world datasets demonstrate the superior performance of our Stable CFR on estimating HTE for underrepresented populations. |
| Researcher Affiliation | Collaboration | 1Department of Computer Science and Technology, Zhejiang University, China 2Key Laboratory for Corneal Diseases Research of Zhejiang Province, Zhejiang University, China 3Department of Quantitative Theory and Methods, Emory University, Atlanta, USA 4School of Economics and Management, Tsinghua University, Beijing, China 5Shanghai AI Laboratory, Shanghai, China 6Shanghai Institute for Advanced Study of Zhejiang University, Shanghai, China. |
| Pseudocode | Yes | Algorithm 1 Epsilon-Greedy Matching with Distance-based Sampling for Point qj. |
| Open Source Code | Yes | The code is available at: https://github.com/anpwu/Stable CFR |
| Open Datasets | Yes | PM-CMR2 (Wyatt et al., 2020) study the impact of PM2.5 partical level on the cardiovascular mortality rate (CMR) in 2132 counties in the US using the data provided by the National Studies on Air Pollution and Health. PM-CMR:https://pasteur.epa.gov/uploads/10.23719/1506014/SES PM25 CMR data.zip |
| Dataset Splits | Yes | we split the samples on each dataset into training/validation data with an 80/20 proportion of training/validation splits. |
| Hardware Specification | Yes | Hardware used: Ubuntu 16.04.5 LTS operating system with 2 * Intel Xeon E5-2678 v3 CPU, 384GB of RAM, and 4 * Ge Force GTX 1080Ti GPU with 44GB of VRAM. |
| Software Dependencies | Yes | Software used: Python 3.7.15 with Tensor Flow 1.15.0, Pytorch 1.7.1, Num Py 1.18.0, and Matplot Lib 3.5.3. |
| Experiment Setup | Yes | We return the best-evaluated iterate on validation data with early stopping and choose the best hyper-parameters from ϵ {0.0, 0.2, 0.4, 0.6, 0.8, 1.0}3 & σ {0.05, 0.1, 0.15, 0.2, 0.25, 0.3}. |