Expectation-Maximization algorithm for bi-variate Normal Inverse Gaussian distribution

Expectation-Maximization (EM) algorithm for bi-variate Normal Inverse Gaussian (biNIG) distribution

Description

EM estimation of parameters of bi variate NIG distribution.

The test file:
1. Simulate biNIG sample with use of randraw.m (http://www.mathworks.com/matlabcentral/fileexchange/7309)
or invgrnd.m (http://www.mathworks.com/matlabcentral/fileexchange/10934) .
2. Calls EMBIVNIG.m (values of starting parameters are chosen arbitrary).
3. Calls binigpp.m for P-P plot to check the fit.

References:

 "EM-estimation and modeling of heavy-tailed processes with the multivariate normal inverse Gaussian distribution", Oigard, Hanssen, Hansen and Godtliebsen, Signal Processing, vol. 85 (2005), p. 1655-1673

"The Two-Dimensional Hyperbolic Distribution and Related Distributions, with an Application to Johannsen's Bean Data", P. Blaesild, Biometrika, vol. 68, No. 1 (Apr., 1981), pp. 251-263, (Theorem 1 (a) & (c), p. 253)

Any comments welcome :)

• Required Products: Statistics Toolbox
• MATLAB release: MATLAB 7.3 (R2006b)

[Zhe]

Thanks for sharing! The plot seemed perfect, but the test result was really bad. Why？

[Zhe]

Accordiong to the paper, line 66 in the function EMBIVNIG should be A = betaem(k+1,:)'*betaem(k+1,:)*ksi.
Is that ?