Computationally efficient method to estimate orthant probabilities of high-dimensional Gaussian vectors. Further implements a function to compute conservative estimates of excursion sets under Gaussian random field priors.
anMC
is a R package to efficiently compute orthant probabilities of high-dimensional Gaussian vectors. The method is applied to compute conservative estimates of excursion sets of functions under Gaussian random field priors. This is an upgrade on the previously existent package ConservativeEstimates. See the paper Azzimonti, D. and Ginsbourger D. (2018) for more details.
The package main functions are:
ProbaMax
: the main function for high dimensional othant probabilities. Computes P(max X > t), where X is a Gaussian vector and t is the selected threshold. The function computes the probability with the decomposition explained here. It implements both the GMC
and GANMC
algorithms. It allows user-defined functions for the core probability estimate (defaults to pmvnorm
of the package mvtnorm
) and the truncated normal sampler (defaults to trmvrnorm_rej_cpp
) required in the method.
ProbaMin
: analogous of ProbaMax
but used to compute P(min X < t), where X is a Gaussian vector and t is the selected threshold. This function computes the probability with the decomposition explained here. It implements both the GMC
and GANMC
algorithms.
conservativeEstimate
: the main function for conservative estimates computation. Requires the mean and covariance of the posterior field at a discretization design.
To install the latest version of the package run the following code from a R console:
if (!require("devtools"))install.packages("devtools")devtools::install_github("dazzimonti/anMC")
Azzimonti, D. and Ginsbourger, D. (2018). Estimating orthant probabilities of high dimensional Gaussian vectors with an application to set estimation. Journal of Computational and Graphical Statistics, 27(2), 255-267. DOI: 10.1080/10618600.2017.1360781. Preprint at hal-01289126
Azzimonti, D. (2016). Contributions to Bayesian set estimation relying on random field priors. PhD thesis, University of Bern. Available at link
Fixed an issue in trmvrnorm_rej_cpp
that could prevent a call to the function conservativeEstimate
from inside another package.
updated references.
The package requires a C++11 compiler.
The time measurements now are taken with the chrono
library from C++11.
Dropped dependency on microbenchmark
package.
The function selectEq
is now called selectActiveDims
The variable Thresh
is now called threshold
The functions ProbaMax
and ProbaMin
now have a user option to choose the function trmvrnorm
to generate truncated multivariate normal samples. The default option is trmvrnorm_rej_cpp
.