Dependence of variance on covariate design in nonparametric link regression

TL;DR

This paper investigates how the variance in nonparametric link regression depends on the covariate design, affecting the prediction of mean outcomes at links between nodes.

Contribution

It highlights the design-dependent nature of variance in nonparametric link regression, a novel insight for modeling link-based outcomes.

Findings

Variance in link regression varies with covariate design

Implications for improving prediction accuracy in network data

Provides theoretical understanding of variance behavior

Abstract

This paper discusses a design-dependent nature of variance in nonparametric link regression aiming at predicting a mean outcome at a link, i.e., a pair of nodes, based on currently observed data comprising covariates at nodes and outcomes at links.