Estimating Joint Treatment Effects by Combining Multiple Experiments
Authors: Yonghan Jung, Jin Tian, Elias Bareinboim
ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Finally, we perform simulation studies, which corroborate the effectiveness of the proposed methods. and Our experimental results corroborate theories. |
| Researcher Affiliation | Academia | 1Purdue University 2Iowa State Univerity 3Columbia University. |
| Pseudocode | No | The paper describes its estimators and procedures through mathematical definitions and textual descriptions, but it does not include any clearly labeled pseudocode blocks or algorithm listings. |
| Open Source Code | No | The paper states 'We implemented the model using Python' and provides code for data generation in Appendix E, but it does not include an explicit statement or link indicating that the source code for the proposed methodology (estimators) is publicly available. |
| Open Datasets | Yes | We applied the proposed estimators to Project STAR dataset (Krueger & Whitmore, 2001; Schanzenbach, 2006). ... We obtained the Project STAR dataset from the following R-package, https://rdrr.io/cran/AER/man/STAR. html. |
| Dataset Splits | No | The paper mentions splitting the dataset into 'training and test samples with a 5:5 ratio' in Section 5.1 and Appendix E, and states that 'Samples for training nuisances and evaluating the estimators equipped with the trained nuisance are separate and independent' (Assumption 2). It also refers to 'cross-fitting algorithms'. However, it does not explicitly specify a distinct validation dataset split or its proportion. |
| Hardware Specification | No | The paper states it used 'XGBoost' and 'implemented the model using Python' but does not specify any hardware details such as GPU/CPU models, memory, or cloud computing resources used for the experiments. |
| Software Dependencies | No | The paper mentions using 'XGBoost' and 'implemented the model using Python' (Appendix E) but does not provide specific version numbers for these software components or any other key libraries. |
| Experiment Setup | Yes | We ran 100 simulations for each N = {2000, 4000, 6000, 8000, 10000} where N is the sample size. We measure the AAEest for each four scenarios: (Case 1)... (Case 2)... (Case 3)... (Case 4)... In modeling nuisance using the XGBoost, we used the command xgboost.XGBClassifier(eval_metric= logloss ) to use the XGBoost with the default parameter settings. |