Abstract

Big datasets found in statistical practice often have a rich structure. Most traditional methods, including their modern counterparts, fail to efficiently use the information contained in them. In this talk I propose and discuss an alternative modelling strategy based on herds of simple models.