interpolate_marginal_posterior: Interpolate the Marginal Posterior

In aghq: Adaptive Gauss Hermite Quadrature for Bayesian Inference

Interpolate the Marginal Posterior

Description

Build a Lagrange polynomial interpolant of the marginal posterior, for plotting and for computing quantiles.

Usage

interpolate_marginal_posterior(
  margpost,
  method = c("auto", "polynomial", "spline")
)

Arguments

margpost

The output of aghq::marginal_posterior. See the documentation for that function.

method

The method to use. Default is a k point polynomial interpolant using polynom::poly.calc(). This has been observed to result in unstable behaviour for larger numbers of quadrature points k, which is of course undesirable. If k > 3, you can set method = 'spline' to use splines::interpSpline() instead, which uses cubic B-Splines. These should always be better than a straight polynomial, except don't work when k < 4 which is why they aren't the default. If you try and set method = 'spline' with k < 4 it will be changed back to polynomial, with a warning.

Value

A function of theta which computes the log interpolated normalized marginal posterior.

See Also

Other summaries: compute_pdf_and_cdf(), compute_quantiles(), marginal_posterior()

aghq documentation built on June 7, 2023, 5:10 p.m.