Approximate Bayesian Inference via Sparse Grid Quadrature Evaluation (BISQuE) for Hierarchical Models
Create a centered and scaled sparse integration grid
Evaluate a mixture density
Compute expectations via weighted mixtures
Named inverse transformation functions
Jacobian for exponential transform
Jacobian for logit transform
Jacobian for log transform
Jacobian for logit transform
Use sparse grid quadrature techniques to integrate (unnormalized) dens...
Wrapper to abstractly evaluate log-Jacobian functions for transforms
Merge pre-computed components of f(theta1 | theta2, X)
Fit a spatially mean-zero spatial Gaussian process model
Draw posterior predictive samples from a spatial Gaussian process mode...
Named transformation functions
Derive parameters for building integration grids
Construct a weighted mixture object
Implementation of the 'bisque' strategy for approximate Bayesian posterior inference. See Hewitt and Hoeting (2019) <arXiv:1904.07270> for complete details. 'bisque' combines conditioning with sparse grid quadrature rules to approximate marginal posterior quantities of hierarchical Bayesian models. The resulting approximations are computationally efficient for many hierarchical Bayesian models. The 'bisque' package allows approximate posterior inference for custom models; users only need to specify the conditional densities required for the approximation.