Segregated Graphs and Marginals of Chain Graph Models

Authors: Ilya Shpitser

NeurIPS 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We illustrate the utility of segregated graphs for analyzing outcome interference in causal inference via simulated datasets.
Researcher Affiliation Academia Ilya Shpitser Department of Computer Science Johns Hopkins University ilyas@cs.jhu.edu
Pseudocode No The paper does not contain any structured pseudocode or algorithm blocks.
Open Source Code No The paper does not provide any concrete access information (e.g., a link or explicit statement) to open-source code for the described methodology.
Open Datasets No The paper states: 'We generated 1000 members of our family described above, used each member to generate 5000 samples'. It describes a data generation process but does not provide access to a publicly available or open dataset, nor a link to the generated data.
Dataset Splits No The paper mentions generating samples and fitting models but does not specify dataset splits for training, validation, or testing.
Hardware Specification No The paper does not provide specific details about the hardware used to run the experiments (e.g., GPU/CPU models, memory, or cloud resources).
Software Dependencies No The paper mentions using 'an approach described in [5]' for fitting the model, but it does not list specific software names with version numbers for reproducibility.
Experiment Setup Yes In all members of this family, A was assigned via a fair coin, p(W | A, U) was a logistic model with no interactions, B1 was randomly assigned via a fair coin given no complications (W = 1), otherwise B1 was heavily weighted (0.8 probability) towards treatment assignment. The distribution p(Y1, Y2 | U, B1, A) was obtained from a joint distribution p(Y1, Y2, U, B1, A) in a log-linear model of an undirected graph G of the form: 1 C( 1) x C 1λC , where C ranges over all cliques in G, . 1 is the L1-norm, λC are interactions parameters, and Z is a normalizing constant. In our case G was an undirected graph over A, B1, U, Y1, Y2 where edges from Y2 to B1 and U were missing, and all other edges were present. Parameters λC were generated from N(0, 0.3).