Description Details Author(s) References See Also
This package gives (a parallel) implementation of the improved Laplace approximation for multivariate integrals of user-written unimodal functions (see Ruli et al. 2015). The method essentially approximates the target integral by the ratio of the (unnormalised) integrand and an approximation of its normalised version, both evaluated at the modal value. The normalised integrand is obtained through a sequential application of the Laplace approximation for marginal densities. Like the standard Laplace approximation, the improved Laplace approximation is a deterministic method which approximates intractable multidimensional integrals by (essentially) numerical optimisations. However, whith respect to the Laplace approximation, the improved Laplace involves scalar numerical integrations. Nevertheless, the improved Laplace approximation tends to be fast and extremely accurate, especially with skewed fat-tailed integrands.
Currently the packages provides two functions iLap
and ilap2d
which perform approximation of d-variate and bivariate integrals, respectively.
Erlis Ruli erlisr@yahoo.it
Ruli E., Sartori N. & Ventura L. (2015) Improved Laplace approximation for marignal likelihoods. http://arxiv.org/abs/1502.06440
iLap
and iLap2d
for more details and examples.
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.