A Comparison of Peer Influence Estimates from SIENA Stochastic Actor–based Models and from Conventional Regression Approaches

Abstract

The current study compares estimates of peer influence from an analytic approach that explicitly address network processes with those from traditional approaches that do not. Using longitudinal network data from the PROmoting School–community–university Partnerships to Enhance Resilience peers project, we obtain estimates of social influence on multiple outcomes from both conventional linear modeling approaches and the stochastic actor–based modeling approach of the simulation investigation for empirical network analysis (SIENA) software. Our findings indicate that peer influence estimates from SIENA are not more conservative relative to other methods, that each method is subject to omitted variable bias from stable individual differences, and that imprecision among each method could lead to erroneous conclusions in samples that lack sufficient power. Together, these results underscore the difficulty in obtaining estimates of peer influence from observational data, and they give no indication that results from conventional methods tend to be biased toward overestimating peer influence, relative to SIENA.

Cognitive Authority and the Constraint of Attitude Change in Groups

“… 12. For models comparing clustered standard errors and more computationally intensive approaches to relational processes and social influence, see Lindgren (2010) and Ragan and colleagues (2019). These offer broad support for the robustness of models presented here. …”

Peer effects in smoking cessation: An instrumental variables analysis of a worksite intervention in Thailand

“…A recent review by Advani and Malde (2018) and work by Carrell et al (2013) points out how not accounting for such heterogenous characteristics within groups may lead to biased estimates of peer effects in traditional models. Despite these challenges, there is recent work showing that traditional regression-based approaches to estimating peer influence can perform well relative to newer, less-restrictive analytic techniques that jointly model network dynamics and selection processes ( Ragan et al, 2019 ).…”

Statistical methods for the estimation of contagion effects in human disease and health networks

“…The model is also relatively inflexible as researchers can only use pre-defined network and behavioral terms. More importantly, although SAOM simultaneously models contagion and network-selection processes, it can still suffer from the aforementioned omitted variable bias problem, and contagion estimates from SAOMs are not more conservative relative to other conventional methods [45] . As Steglich et al pointed out, estimates from SAOMs are still biased when “non-observed variables co-determine the probabilities of change in network and/or behavior [40] .”…”