etpinard / gaussiandistributions.jl

Gaussian distributions as state variables and for uncertainty quantification with Unitful

GaussianDistributions

To install run Pkg.add("GaussianDistributions").

This package creates a simplistic alternative to Distributions to generate Gaussian or normally distributed random variables and deal with their distributions.

The raison d’être of GaussianDistributions is that the type hierarchy layed out by distributions is not well suited to handle singular covariance matrices, and anything other than Float64s and Vector{Float64}s.

This package contains enough functionality such that Gaussians can be useful as Gaussian state variable for example in a Kalman filter.

It also provides cumulative distributions functions (cdf) for bivariate Normal distributions using an approximation (valid unless the correlation ρ is very strong and x ≈ y*sign(ρ).)