Minority Representation and Relative Ranking in Sampling Attributed Networks

We explore two questions related to sampling and minorities in attributed networks with homophily. The first question is to investigate sampling schemes which favor minority attribute nodes and which give preference to “more popular” nodes having higher centrality measures in the network. A data study shows the efficiency of Page-rank and walk-based network sampling schemes on a directed network model and a real-world network with small minorities. The second question concerns the effect of homophily and out-degrees of nodes on the relative ranking of minorities compared to majorities in degree-based sampling. Several synthetic network configurations are considered and the conditions for minority nodes to have a higher relative rank are investigated numerically. The results are also assessed with real-world networks.