Univariate Multimodal Random Number Generator

Generate random numbers from a mixture of Gaussian distributions.

Sometimes we need samples that come from a mixture of pdf's, where more than 1 peaks are observed. E.g. for analyzing goodness of a non-parametric kernel density estimation method.

By using the Rejection method, this function generates random numbers from a mixture (average sum) of N normal (Gaussian) distributions. It gives also the mixture probability density function (pdf) as an optional output.

An example of 500 samples generated from 3 Gaussian distributions are shown in the figure, i.e. by the following command:

>> X = umgrn([-4 0 5],[1 2 1.5],500);
>> figure; hist(X);

See 'help umgrn' for more details.

Authors:
1. Avan Suinesiaputra ([e-mail])
2. Fadillah Tala ([e-mail])