GFBBIS

GFBBIS is an implementation of gradient and gradient free black box importance sampling originally from Qiang Liu, & Jason D. Lee. (2016). Black-box Importance Sampling. https://arxiv.org/abs/1610.05247

Installation

You can install the development version of GFBBIS like so:

devtools::install_github("travis-j-hahn/GFBBIS")

Example

Here is the basic syntax for the gradient version of black box importance sampling:

library(GFBBIS)
## basic example code
out = BBIS(theta,theta_grads,1000,kernel='rbf')

print(out$adj_mean)

MSEs of the different forms can be seen here:

A full example can be found in examples/logit_example.R