Senior Honors Thesis and Final ARTU Presentation: “Exponential Random Graph Models Under Measurement Error”

Abstract:  Understanding social networks is increasingly important in a world dominated by social media and access to enormous amounts of data. When analyzing social network data, we are often interested in the underlying structures that exist in the network. Exponential random graph models (ERGMs) are frequently used by analysts to gain a better understanding of the formation and operation of these structures. Data collection, however, always involves measurement error that causes the observed network to differ from the true network. This, in turn, introduces error into the resulting fitted model. In our study, we introduced simulated measurement error into a social network of high school friendships in order to investigate the robustness of ERGMs when faced with noisy data. The resulting ERG coefficients and descriptive statistics of the perturbed networks were compared to the original, unperturbed values.

Host: Nan Lin