Getting the normalizing constant by integration: positive infinity doesn't work

[bbbales2]

The Boost integrators we used are doc’d here: Chapter 13. Quadrature and Differentiation - 1.75.0

There’s interesting stuff in there for sure. For instance there’s this on endpoints: Handling functions with large features near an endpoint with tanh-sinh quadrature - 1.75.0, but it sounds like ideally your quadrature goes to infinity anyway.

I wonder if there’s a way to use arbitrary precision math with a quadrature package in R: https://cran.r-project.org/web/packages/Rmpfr/vignettes/Rmpfr-pkg.pdf ?

@martinmodrak

But one does not need to use such extreme numbers to see small discrepancies

Interesting example!

@bnicenboim

I am using cmdstan 2.26 though.

You should do another install_cmdstan() and upgrade to 2.26.1 but that won’t address this (Release of CmdStan 2.26.1).