Local Linear Forests
This method addresses the problem of poor predictive performance in non-parametric regression for researchers and practitioners dealing with smooth effects, representing a novel improvement rather than an incremental change.
The paper tackles the limitation of random forests in fitting smooth signals by introducing local linear forests, which pair the forest kernel with a local linear regression adjustment, resulting in improved asymptotic convergence rates and substantial accuracy gains on real and simulated data.
Random forests are a powerful method for non-parametric regression, but are limited in their ability to fit smooth signals, and can show poor predictive performance in the presence of strong, smooth effects. Taking the perspective of random forests as an adaptive kernel method, we pair the forest kernel with a local linear regression adjustment to better capture smoothness. The resulting procedure, local linear forests, enables us to improve on asymptotic rates of convergence for random forests with smooth signals, and provides substantial gains in accuracy on both real and simulated data. We prove a central limit theorem valid under regularity conditions on the forest and smoothness constraints, and propose a computationally efficient construction for confidence intervals. Moving to a causal inference application, we discuss the merits of local regression adjustments for heterogeneous treatment effect estimation, and give an example on a dataset exploring the effect word choice has on attitudes to the social safety net. Last, we include simulation results on real and generated data.