METHOD NON-GAUSSIAN DATA DEFINITION BY PROBABILISTIC MODELS BASED ON THE TRUNCATED NORMAL DISTRIBUTIONS

I.K. Vasilyeva

The statistical model for the components of multidimensional random variable based on the truncated normal distributions and the technique of the models parameters estimation are proposed. The results of the approximation a number of non-Gaussian distributions by the truncated Gaussian probability distributions and their mixtures are produced. Test of the models acceptability completed for to describe the results of random variables simulation with uniform, exponential, Rayleigh and arcsine probability distributions. It is shown that the proposed model is adequate, thereby justified background for the development of a sufficiently universal multidimensional statistical model in the form of K-components mixture of N-dimensional truncated Gaussian distributions.

 

Key words: truncated normal distribution, distribution parameters, mixture parameters, statistical estimation, criterion function, approximation.