Jackknife - after - Bootstrap: Detecting Influential Actors in a Network.

Google map

With the increasing prevalence of social networks in modern science, it is important to identify actors that influence the flow of information. We propose a non-parametric method for detecting influential actors in a network by adapting Jackknife-after-Bootstrap (JaB) algorithm for detecting influential points in regression. We define influential actors as those that are abnormally central compared to the others in the network. This algorithm allows us to (1) provide a list of which actors are influential and which are not, (2) quantify how close each actor is to being flagged as influential, and (3) rank all actors from most to least influential. JaB for network data can incorporate any node-specific centrality statistic to account for various types of information flow. We apply our method to various investigative applications to identify persons of interest.

Olivia Beck is a 4th-year PhD student in the Department of Statistics with a dual title in Social Data Analytics. She graduated in 2020 from Colorado State University with a B.S. in Applied Mathematics and a B.S. in Statistics. Her research interests include empirical likelihood and network models with social data applications.