multinor

A succesful implementation of numerical computation of multivariate normal probabilities.

A collection of M-Files that provide a succesful implementation of the difficult problem of numerical computation of multivariate normal probabilities. This is very useful for the distributed detection with nuisance parameters and in classical multivariate problems.

Two methods are used for which references are indicated. Function mp.m uses thetrachoric series (see K. Pearson algorithm) to compute simple integrals, from minus infinity to the upper limit of integration (dimension is n) when the covariance has 1 on the diagonal.

Function genz.m uses some linear transformation of the lower and upper bounds and the Monte Carlo integration (see A. Genz algorithm) and is more precise. Any covariance matrix (obviously definite positive) can be used. When increasing accuracy, the computation time grow.

A demo (demomp.m) demonstrate the use of these functions. The function intgc.m is used by genz.m

Algorithms were tested under Matlab 5.2