Assessing the Relational Outcome Model through Monte Carlo Simulations

• In Event: 40221 - Section on Mathematical Sociology Flash Talks
  In Refereed Roundtable (90 minute): Table 1: Assessing the Relational Outcome Model through Monte Carlo Simulations

Abstract

Inferential network analysis techniques are specifically designed to model observations that are autocorrelated in sociometric (i.e., network) space, a feature that traditional statistical approaches cannot effectively deal with. Importantly, inferential network analysis techniques are specifically designed to model either the networks themselves, endogenous inferential network analysis, or outcomes exogenous to the networks, exogenous inferential network analysis. Very recent research has started to leverage the relational event model—which was specifically designed to model time-stamped relational event sequences as endogenous outcomes—to model exogenous outcomes via the relational outcome model. In particular, the relational outcome model is a novel technique, specifically designed to model exogenous outcomes resulting from time-stamped relational event sequences. This paper aims to assess how the relational outcome model is superior to traditional statistical models, namely, the ordinary least squares estimator, in reducing bias, inefficiency, and incorrect inference for exogenous
covariates when outcomes are interdependent across relational outcomes. Following prior research on the network autocorrelation model and the quadratic assignment procedure, we employ a set of Monte Carlo simulation procedures to examine two statistical properties of the relational outcome model in comparison to the traditional ordinary least squares estimator: 1) the rate of bias in an exogenous covariate when the outcomes are correlated across a relational sequence, and 2) the efficiency of parameter estimates when the outcomes are correlated with the past events (i.e., past events sharing the same actors).

Authors

• Kevin Anthony Carson, University of Arizona

• Diego F. Leal, University of Arizona