Research

"Messy Data, Robust Inference? Navigating Obstacles to Inference with bigKRLS" with Robert Shaffer was accepted for publication by Political Analysis on May 18, 2018. In addition to extending KRLS computationally, the paper proposes a p-value correction and demonstrates KRLS' robustness and nuance through a variety of cross-validations and simulations in the context of the 2016 US presidential election (Paper, Project Overview, Slides, Video [International Methods Colloquium]). Robert and I are also excited to start a new collaboration with Chad Hazlett, Luke Sonnet, Arash Amini, and Carlos Cinelli, which is supported by the UCLA Center for Social Statistics and that aims to extend KRLS to even bigger data and a wider set of methodological circumstances.

Despite their growing ubiquity (of which I suppose some of my software is now a part), neural nets should not be considered a 'go to' option for prediction. Xi Cheng, Bohdan Khomtchouk, Norm Matloff and I provide theoretical reason to expect polynomial regression may often be a convenient alternative to NNs and show that on a wide variety of behavioral and biological data sets, polynomial regression performs at least as well; arXiv:1806.06850.

Through the Center for Deliberative Democracy at Stanford, Robert Luskin and I are studying how prior convictions towards the European Union and immigration are changed by small group discussion with people from many different countries.

Terri Givens, John Graeber, and I are researching the politics of whether European cabinets focus on immigration control or immigrant integration. Along with Rachel Navarre, the three of us are also working on a textbook on comparative immigration politics for Routledge.