Title: Optimal design of experiments for networked data
Authors: Ben Parker - University of Southampton (United Kingdom) [presenting]
Abstract: The problem of how to optimally design experiments for a large number of participants who are connected by a network structure is considered. As a motivating example, consider determining which of a selection of adverts on Facebook is effective, given that users have some friendship structure, as well as potentially some other characteristics which can be expressed as a blocking structure. We introduced previously a linear network effects model, and found optimal designs when experimental units were connected according to some relationship, which was specified by an adjacency matrix. We showed how networks could be useful in a variety of applications: for example, agricultural experiments where experimental units (plots) were connected by some spatial relationship, and also in crossover trials, where experimental units were connected by temporal networks. We develop now faster algorithms that allow optimal designs on networks to be found more quickly, and makes the problem of finding optimal designs for large networks manageable. We also argue that there is a wide class of experiments that can be reformulated into a problem of design on a network. By regarding experimental design as a problem in network science, we can improve experimental design algorithms for large networks, and also find designs even when there is no obvious network relationship.