Active Adaptive Experimental Design for Treatment Effect Estimation with Covariate Choice
Authors: Masahiro Kato, Akihiro Oga, Wataru Komatsubara, Ryo Inokuchi
ICML 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | This section presents simulation studies. There are two objectives for the simulation. First, through simulation studies, we clarify the problem setting. Second, for the problem, we confirm the soundness of our designed experiment. ... In each of our simulations, we conduct 200 experiments and compute the performance measures. In this section, in the y-axis of Figures 6 7, we report the empirical MSEs for θ0; that is, 1 200 P200 i=1(θ0 ˆθ(i) T )2, where ˆθ(i) T denotes the estimator of ATE in the i-th trial. |
| Researcher Affiliation | Industry | 1Mizuho-DL Financial Technology Co., Ltd., Tokyo, Japan. Correspondence to: Masahiro Kato <masahirokato@fintec.co.jp>. |
| Pseudocode | Yes | We describe the details below, and its pseudo-code is provided in Algorithm 1. |
| Open Source Code | No | The paper does not provide any explicit statements or links indicating that the source code for the described methodology is publicly available. |
| Open Datasets | Yes | This section provides simulation studies using the semi-synthetic dataset called the Infant Health and Development Program (IHDP). The IHDP dataset has simulated outcomes and real-world covariates (Hill, 2011). |
| Dataset Splits | No | The paper describes the data generation process for its simulation studies (e.g., "In each round t, we sample covariate Xt with replacement from the 747 samples"), but it does not specify explicit training, validation, or test dataset splits using percentages, counts, or citations to predefined splits. |
| Hardware Specification | Yes | The simulation studies were conducted by using the Mac Book Pro and a Linux ubuntu18 server with 64 cores and 251 GB of RAM. |
| Software Dependencies | No | The paper does not specify any software dependencies with version numbers (e.g., specific Python libraries or frameworks with their versions) that would be needed for reproducibility. |
| Experiment Setup | Yes | Basic setup Let T = 2000. We specify the means as EX q(x)[Y (1)] = 10 and EX q(x)[Y (0)] = 7, under which the ATE is θ0 = 3. Let covariates x be one-dimensional. For any x, the conditional means are µ0(1)(x) := C1 x + 3x2 1 , µ0(0)(x) := C0 (0.1x + 0.2) , where C1 and C0 are parameters so that EX q(x)[µ0(1)(X)] = EX q(x)[Y (1)] = 10 and EX q(x)[µ0(0)(X)] = EX q(x)[Y (0)] = 7. ... σ2 0(1)(x) := 2 + 1.2sin(2x) + (x + x2)/25, σ2 0(0)(x) := 2 + 0.8cos(x/2) + x2/50. |