The Director of CESS, Professor Raymond Duch, presented his paper entitled “Enhancing experimental design with multiple modes, machine learning and digital trace” at Oxera Consulting LLP, in London, in 30th January.
“An important challenge for social scientists is establishing the robustness and generalizability of estimated treatment effects. In a series of recent publications, we have proposed novel design and estimation strategies that help us move in this direction. These relatively well-known methods include machine learning, multiple regression post-stratification (MRP), and multi-mode design. The novelty is implementing them in the context of large-scale experiments, particularly in an online environment. The presentation provides an overview of these methods along with applied illustrations from a number of our recent experimental projects.”