Multi-Normal Distribution

Lapsilago · October 12th, 2006, 7:12 am

Hi All,I am looking for algorithms efficiently computing a multivariate normal distribution withgiven mean vector and covariance/correlation matrix.Anyone any good references?Best wishes, Lapsi

DavidJN · October 12th, 2006, 3:27 pm

http://www.math.wsu.edu/math/faculty/genz/homepage

SU2 · October 14th, 2006, 8:16 pm

Assuming you know how to simulate uncorrelated standard normals Z_1 , ... , Z_n. Then (using Einstein summation) the vectors U_i = D_ij Z_j , i=1,...,n, are normally distributed with covariance matrix D'D (where D' = D transpose). In fact, D is just the Cholesky decomposition of the covariance matrix.

Lapsilago · October 23rd, 2006, 10:42 am

Dear all,recently found a paper on the multinormal distribution. It gives an algorithm for computing probabilities,i.e. computing values of the distribution function. It is:Schervish, M. "Algorithm AS195:Multivariate Normal Probabilities With Error Bound" Applied Statistics, 34, (Jan 1985), 103-104Best, Lapsi