Multivariate random distribution

[meyerma]

Hi @moneta

Is there any one-liner multivariate normal distribution function taking a covariant TMatrix as input and generating a column vector (such as TVector) of N dimension?

Something analog to:

scipy.stats.multivariate_normal — SciPy v1.16.0 Manual

Thank you, M.

[ferhue]

There is:

ROOT: tutorials/roofit/roostats/MultivariateGaussianTest.C Source File

or for 2D:

Tilt of ellipse in bigaus

The “bigaus” is actually based on the “ROOT::Math::bigaussian_pdf” function (you can use the “rho”, “sigmax” and “sigmay” parameters to calculate the principal axes, which are defined by the eigenvectors of the covariance matrix, and the semidiameters, which are the square-roots of its eigenvalues): { // ... create the function ... TF2 *f = new TF2("f", "[0] * ROOT::Math::bigaussian_pdf(x, y, [2], [4], [5], [1], [3])", -10., 10., -10., 10.); f->SetParNames("constant", "x0", "sigmax", "y0"…

so not really a onliner solution yet, but PRs are welcome

[meyerma]

I am going to check that. Thanks for sharing the test!