Orthogonal Random Forest for Causal Inference
Authors: Miruna Oprescu, Vasilis Syrgkanis, Zhiwei Steven Wu
ICML 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We provide a comprehensive empirical evaluation of our algorithm on both synthetic and real data. and 7. Monte Carlo Experiments We compare the empirical performance of ORF with other methods in the literature (and their variants). |
| Researcher Affiliation | Collaboration | Miruna Oprescu 1 Vasilis Syrgkanis 1 Zhiwei Steven Wu 2 *Equal contribution 1Microsoft Research New England 2University of Minnesota Twin Cities. Correspondence to: Miruna Oprescu <moprescu@microsoft.com >, Vasilis Syrgkanis <vasy@microsoft.com>, Zhiwei Steven Wu <zsw@umn.edu>. |
| Pseudocode | No | The paper describes the Orthogonal Random Forest algorithm in Section 3 ('Orthogonal Random Forest') but does not provide it in a structured pseudocode or algorithm block format. |
| Open Source Code | Yes | The source code for running these experiments is available in the git repo Microsoft/Econ ML. |
| Open Datasets | Yes | We apply our method to estimate heterogeneous treatment effects from observational data with discrete treatments or continuous treatments... Finally, to motivate the usage of the ORF, we applied our technique to Dominick s dataset, a popular historical dataset of store-level orange juice prices and sales provided by University of Chicago Booth School of Business. and The data generating process we consider is described by the following equations: Yi = 0(xi) Ti + h Wi, γ0i + "i, Ti = h Wi, β0i + i. Moreover, xi is drawn from the uniform distribution U[0, 1], Wi is drawn from N(0, Ip)... |
| Dataset Splits | No | The paper describes a Monte Carlo simulation setup where data is generated and evaluated at 'test points', and mentions 'cross-validated Lasso' for parameter selection, but it does not provide specific training, validation, and test dataset split percentages or sample counts in the main text. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., CPU/GPU models, memory) used for running its experiments. |
| Software Dependencies | No | The paper mentions using the 'GRF R package' but does not specify its version number or any other software dependencies with explicit version information. |
| Experiment Setup | Yes | We implement ORF as described in Section 3, setting parameters under the guidance of our theoretical result: subsample size s (n/ log(p))1/(2 +1), Lasso regularization λγ, λq log(p)s/n/20 (for both tree learner and kernel estimation), number of trees B = 100 n/s, a max tree depth of 20, and a minimum leaf size of r = 5. |