Optimal Model Averaging of Mixed-Data Kernel-Weighted Spline Regressions

Model averaging has a rich history dating from its use for combining forecasts time-series models (Bates and Granger) presents compelling alternative to model selection methods. We propose frequentist procedure defined over categorical regression splines (Ma, Racine, Yang) that allows mixed-data predictors, as well nonnested heteroscedastic candidate models. demonstrate the asymptotic optimality of proposed estimator, develop post-averaging inference theory it. Theoretical underpinnings are provided, finite-sample performance is evaluated, an empirical illustration reveals method capable outperforming range popular criteria in applied settings. An R package available practitioners (Racine).