gencov

Generate random covariance matrices, and draw MVN samples using them.

Covariance matrix:

The genS and genArray functions produce random covariance matrices (as ndarray or javascript array) with a specified variance structure. The eigenvalues (principal component variances) V for the covariance matrix may be specified, or may be randomly generated from within a specified range. A random orthogonal matrix Q is generated and its columns used as eigenvectors. The covariance matrix is then generated as S = Q V Q~

Sampling:

Given a covariance ndarray S, you can generate samples from the associated multivariate normal distribution using the mvnrnd function (which creates a function that draws samples from N(mean, S))

Samples x ~ N(0, S) are drawn by first drawing z ~ N(0, I) then transforming x = L z, where S = L L~.

Example usage:

var gencov = require('gencov');
 
// generate a 3-d correlation matrix with variances between 1 and 10,
// and return it as an ndarray:
 
var S = gencov.genS(3);
 
// generate a 5-d correlation matrix with principal components,
// return as a regular array
 
var S = gencov.genArray([3, 2, 1, 0.5, 0.1]);
 
// draw 10 3d samples from a N([a,b,c], S) distribution with random S,
// return as an array of 3-vectors.
 
var X = Array.apply(null, 10).map(mvnrnd([a,b,c], genS(3)))

dependencies

ndarray: https://www.npmjs.com/package/ndarray ndarray-blas-level1: https://www.npmjs.com/package/ndarray-blas-level1 ndarray-blas-level2: https://www.npmjs.com/package/ndarray-blas-level2 ndarray-blas-dger: https://www.npmjs.com/package/ndarray-blas-dger ndarray-unpack: https://www.npmjs.com/package/ndarray-unpack ndarray-gram-schmidt-qr: https://www.npmjs.com/package/ndarray-gram-schmidt-qr ndarray-cholesky-factorization: https://www.npmjs.com/package/ndarray-cholesky-factorization