doit: doit: A package for approximate Bayesian computation using...

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

Description

The package uses as input a data frame of parameter values theta and corresponding evaluations of an unnormalised probability density function f(theta), as is a common situation in Bayesian inference problems. By interpolating the target density using Gaussian kernels, the normalisation constant, as well as marginal densities and expectations can be approximated.

Fitting methods

doit_estimate_w obtains the optimal width of the Gaussian interpolation kernels by cross validation.

doit_fit estimates the parameters for the DoIt approximation

doit_update updates a fitted DoIt approximation by a new parameter value and function evaluation.

Approximation methods

doit_approx evaluates the DoIt approximation at different parameter values

doit_marginal approximates the marginal distribution

doit_marginal_A approximates the marginal distribution of a linear transformation of the inputs

doit_expectation and doit_variance approximate expectation and variance

doit_integral approximates the integral under the unnormalised density

sieste/doit documentation built on May 9, 2019, 4:10 p.m.