sieste/doit: Bayesian Computation using Design of Experiments-Based Interpolation Technique

This package is an implementation of the design of experiments-based interpolation technique (DoIt) for approximate Bayesian computations. The method uses evaluations of the unnormalised posterior density at a space-filling design of parameter values. Normalisation is achieved by approximating the posterior density by a weighted sum of Gaussian kernels. DoIt allows for approximate calculation of marginal posterior densities, and posterior expecations and variances. The package contains functions to optimally choose additional design points, and to calculate the optimal kernel bandwith by efficient cross validation.

install.packages("remotes")
remotes::install_github("sieste/doit")