Gaussian process priors and dependent Dirichlet processes
Michael J. Daniels, Antonio Linero, Jason A. Roy · 2023
We introduce Gaussian processes (GPs) as an alternative to Bayesian additive regression trees as priors on unknown functions. We then show how GPs can be combined with Dirichlet process mixtures to construct dependent Dirichlet process (DDP) mixture models, which can be used for various causal inference problems where priors for unknown conditional densities are needed, particularly for continuous outcomes. We provide details on posterior computations.