Next: A more advance example Up: Models with multivariate Gaussian Previous: Simple types: ex_basic

Another mixture example: ex_bic

ex_bix is a richer example with mixtures where you analyze some simulated data consisting of mixture observations plus a small portion of outliers (simulated from another density). To assess the fit provided by mixture models with different numbers of components, the BIC (Bayesian Information Criterion) is used [5]. The aim of this example is not to promote the use of BIC for selecting the number of component in a mixture model (see the comments in ex_bix concerning the sensitivity of the procedure to various parameters of the analysis) but rather to illustrate the use of H2M functions in a more ``realistic'' setting.



Olivier Cappé, Aug 24 2001