Stephen Loftus

Background

About Stephen Loftus

As a statistical researcher, my time is split between my academic topics and sabermetric topics. From an academic statistics perspective, my research is mostly devoted to variable selection techniques in oversaturated models with specific applications to omics, biological, and ecological datasets. Many techniques designed to be model selection consistent fail in oversaturated models, and as such I worked to develop techniques that accurately identify variables important to a response (Detailed fully in my dissertation). Outside of this, my academic interests include nonparametric penalized splines, data mining techniques, Bayesian sequential updating, and small sample Bayesian posterior estimation. While I keep my sabermetric interests distinct from my academic interests, I believe sabermetric datasets and topics are ripe for further investigation. The questions involved in creating projection systems are ideal for small sample size theory. My specific research interests from a sabermetric perspective include pitch mixing research, opponent adjusted statistics, disrupted pitch sequence analysis, and small sample size research.