Normal variance mixtures: Distribution, density and parameter estimation

Efficient algorithms for computing the distribution function, (log-)density function and estimating parameters of multivariate normal variance mixtures are introduced. For evaluation randomized quasi-Monte Carlo (RQMC) methods utilized in a way that improves upon existing proposed special case t distributions. evaluating log-density an adaptive RQMC algorithm similarly exploits superior convergence properties is This allows parameter estimation task to be accomplished via expectation–maximization-like where all weights log-densities numerically estimated. Numerical examples demonstrate suggested quite fast. Even high dimensions around 1000 can estimated with moderate accuracy using only few seconds run time. Also, even ?100 accurately quickly. An implementation presented available R package nvmix (version ?0.0.4).