Goodness-of-fit testing for behavior in joint dynamic network/behavior models with an extension to two-mode networks

Citation

Wang, Cheng; Butts, Carter T.; Hipp, John R.; Jose, Rupa; & Lakon, Cynthia M. (2017). Goodness-of-fit testing for behavior in joint dynamic network/behavior models with an extension to two-mode networks. First North American Social Networks Conference (NASN2017). Washington, DC.

Abstract

The recent popularity of models that capture the dynamic co-evolution of both network structure and behavior has driven the need for indices to assess the fit of these models. Whereas there are several existing indices for assessing the ability of the model to reproduce the network over time, to date there are few indices for assessing the ability of the model to reproduce the behavior of the individuals in the sample over time. Drawing on the strategy of assessing the fit of a model by comparing global values of the distribution of behavior in the actual observed network to those from networks simulated based on the model parameter values, we propose four goals that a researcher could reasonably expect of a joint structure/behavior model regarding how well it captures behavior, and describe indices for assessing each of these. First, one fundamental goal is that the model should reproduce the distribution of the behavioral variable(s) in the observed sample at each time point. Second, given the dynamic nature of the model, we would expect that a useful model should accurately capture transitions over time in the level of the behavioral variable(s). A third goal for a useful model is that it should generate networks in which the behavior patterns align with key statistics (e.g., vertex-level indices) of the social network. Finally, a fourth goal is that a useful model should be able to capture the degree of behavior clustering within the network. These reasonably simple and easily implemented indices can be used for assessing model adequacy with any dynamic network models jointly working with networks and behavior, including the Stochastic Actor-Based model using the RSiena software package. We demonstrate the indices with an empirical example to show how they can be employed in practical settings and extend to the affiliation dynamics in two-mode networks.

Reference Type

Conference proceeding

Book Title

First North American Social Networks Conference (NASN2017)

Author(s)

Wang, Cheng
Butts, Carter T.
Hipp, John R.
Jose, Rupa
Lakon, Cynthia M.

Year Published

2017

City of Publication

Washington, DC

Reference ID

8232