Inference for Network Regression Models with Community Structure
Authors: Mengjie Pan, Tyler Mccormick, Bailey Fosdick
ICML 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | To evaluate the performance of our proposed blockexchangeable error model, we generate data from a modified latent space model (Hoff, 2005), which satisfies the requirements for block exchangeability. We consider a simple regression model with one covariate, as in (3) where both coefficients equal 1. We consider three settings for the relationship between the covariate and block structure, and three types of covariates. Figure 3 shows the coverage of 95% confidence intervals for β1 for all nine simulation settings. ... We demonstrate our method on data representing passenger volume between US airports (Bureau of Transportation Statistics, 2016). |
| Researcher Affiliation | Collaboration | 1Facebook, Seattle, Washington, USA 2Department of Statistics and Department of Sociology, University of Washington, Seattle, Washington, USA 3Department of Statistics, Colorado State University, Fort Collins, Colorado, USA. |
| Pseudocode | Yes | Algorithm 1 Known block estimation of ΩB... Algorithm 2 Block membership estimation |
| Open Source Code | No | The paper does not contain an explicit statement or link indicating that the source code for the described methodology is publicly available. |
| Open Datasets | Yes | We demonstrate our method on data representing passenger volume between US airports (Bureau of Transportation Statistics, 2016). The number of passenger seats is a right-tailed skewed distribution, so we use the values yij = log(yij + 1) as the relational observations for regression model in (1). For covariates, we calculated the great circle distance between two airports using their longitudes and latitudes. Additionally, we identified the county of the municipality of each airport, and found the total GDP of that county from of Economic Analysis (2015) and average payroll of an employed person from Bureau (2015). |
| Dataset Splits | No | For the simulation studies, the paper describes data generation ('we generate data from a modified latent space model... We consider networks of size 20, 40, 80, and 160') rather than specific train/validation/test splits of a pre-existing dataset. For the Air Traffic Data, it does not explicitly provide details about dataset splits for training or validation. |
| Hardware Specification | No | The paper mentions computational time on a 'standard machine' and 'standard laptop' but does not provide specific hardware details such as CPU/GPU models, memory, or cloud instance types. |
| Software Dependencies | No | The paper mentions using 'optim in R' but does not specify the version numbers for R or the 'optim' package/function. |
| Experiment Setup | Yes | We consider a simple regression model with one covariate, as in (3) where both coefficients equal 1. We generated 1000 errors for each of 500 simulations of the covariates and block memberships, and considered networks of size 20, 40, 80, and 160. ... To numerically optimize the pseudo-likelihood, we used optim in R, with method="L-BFGS-B". We do not set bounds on β, but did place a lower bound of 1e 2 for all variance parameters and a bound of [ 0.9, 0.9] for all correlation parameters. |