Covariate balancing using the integral probability metric for causal inference

Authors: Insung Kong, Yuha Park, Joonhyuk Jung, Kwonsang Lee, Yongdai Kim

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

Reproducibility Variable Result LLM Response
Research Type Experimental We illustrate the superiority of the CBIPM to existing baselines by analyzing simulated and real datasets. In Section 5.1 and 5.2, we present the experimental results using simulation and real datasets respectively.
Researcher Affiliation Academia Insung Kong 1 Yuha Park 1 Joonhyuk Jung 1 Kwonsang Lee 1 Yongdai Kim 1 1Department of Statistics, Seoul National University.
Pseudocode Yes Algorithm 1 Proposed algorithm for the ATT
Open Source Code Yes The code is available at https://github.com/ggong369/CBIPM.
Open Datasets Yes We generate simulated datasets using the Kang-Schafer example (Kang & Schafer, 2007). The Tennessee Student/Teacher Achievement Ratio experiment (STAR) is a 4-year longitudinal class-size study... (Achilles et al., 2008).
Dataset Splits No The paper does not specify exact train/validation/test split percentages or sample counts for any of the datasets used, nor does it refer to predefined splits with citations for these purposes.
Hardware Specification Yes We use R (ver. 4.0.2), Python (ver. 3.6), and NVIDIA TITAN Xp GPUs to obtain the estimates of the ATT and the ATE.
Software Dependencies Yes We use R (ver. 4.0.2), Python (ver. 3.6)... We use twang package (Ridgeway et al., 2017)... CBPS is implemented using CBPS package (Fong et al., 2022)... EB is implemented using EB package (Hainmueller & Hainmueller, 2022)... We use Adam (Kingma & Ba, 2014) optimizer...
Experiment Setup Yes For both the P-CBIPM and the N-CBIPM, we use a neural network with 100 hidden nodes with leaky relu. We use Adam (Kingma & Ba, 2014) optimizer with lr = 0.03 and T = 1000 for gradient descent steps, and Adam optimizer with lradv = 0.3, Tadv = 5 for gradient ascent steps. τ = 0.3 and R = 100 are used.