On the distribution of nutritional status transition probabilities through a hierarchical beta mixture model

This work addresses the problem of estimating the probability density function for each nutritional status considering the sex and the kind of transition a person suffers between two consecutive periods. To this end, an optimization scheme is proposed based on a hierarchical model considering previously known information as input to estimate mixture densities through the expectation maximization algorithm. Besides, a goodness-of-fit metric is shown based on the shared area of the actual density obtained from a Dirichlet distribution and its close relation with the Beta distribution. Thus, several experiments are performed showing a goodness-of-fit of 93.1%, at least.