Ian Barnett

My methodological interests include predictive modeling with machine learning and neural networks, signal detection theory, change point detection, network science, longitudinal data analysis, and multiple testing. In each case I am most excited by problems that deal with correlation. My dissertation thesis, for example, focused on adapting the Higher Criticism, a method used for signal detection, to accomodate the case when the underlying tests are correlated. Just because correlation makes statistical inference difficult is not a sufficient reason to ignore it!

My applied interests lie primarily in mobile health (mHealth) with an emphasis on smartphone-monitoring studies. Smartphone sensor data can be collected passively over long periods of time without any need for incentives and provide a 24/7 view into clinically relevant behaviors such as sleep, physical activity, mobility, and social activity. Most of this work is supported generously by the National Institute of Mental Health (R01MH116884).

I also enjoy sports statistics and do predictive modeling for baseball as a hobby. In particular, I use PitchFX and historical data to predict matchup-specific batter/pitcher outcomes.