Estimating individual treatment effects under unobserved confounding using binary instruments

Authors: Dennis Frauen, Stefan Feuerriegel

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

Reproducibility Variable Result LLM Response
Research Type Experimental Across various computational experiments, we demonstrate empirically that our MRIV-Net achieves state-of-the-art performance.
Researcher Affiliation Academia Dennis Frauen1, 2 & Stefan Feuerriegel1, 2 1 Munich Center for Machine Learning 2 LMU Munich
Pseudocode Yes Algorithm 1: MRIV
Open Source Code Yes Reproducibility: The codes for reproducing the experimental results can be found at https: //github.com/Dennis Frauen/MRIV-Net.
Open Datasets Yes The so-called Oregon health insurance experiment (OHIE) (Finkelstein et al., 2012) was an important RCT with non-compliance. [...] Data available here: https://www.nber.org/programs-projects/projects-and-centers/oregon-health-insurance-experiment
Dataset Splits Yes For all methods except KIV and DFIV, we split the data into a training set (80%) and a validation set (20%).
Hardware Specification Yes For training, we used an AMD Ryzen Pro 7 CPU.
Software Dependencies No The paper mentions using the 'scikit-learn package' and 'R package Biased Urn' but does not provide specific version numbers for these or any other software dependencies such as deep learning frameworks (e.g., PyTorch, TensorFlow) or Python.
Experiment Setup Yes We use two hidden layers with RELU activation functions. We also incorporated a dropout layer for each hidden layer. We trained all models with the Adam optimizer (Kingma & Ba, 2015) using 100 epochs. [...] The tuning ranges for the hyperparameter are shown in Table 6.