What Counts as a Weak Tie? A Comparison of Filtering Techniques to Analyze Co-Exposure Networks

Co-exposure networks are a useful tool to analyze patterns of news consumption. In these networks, nodes are news sources and ties measure the strength of their audience overlap. In observational data, some overlap might result from random exposure or random browsing behavior. Filtering techniques can help eliminate the weakest connections and, with them, random noise. There are different approaches to filtering weighted networks; however, it is not always clear which approach is the most appropriate. Here, we describe three different techniques and we compare their performance using two observed networks. First, we assess the impact that each technique has on the global topology. We then study the trade-off that exists between removing ties and preserving the connectedness of the network, and we offer a mathematical approach to systematically analyze that trade-off. We propose a method to identify an optimal threshold to maximize the number of edges removed while minimizing the number of nodes that become isolates.