Mostratge i estimació de poblacions ocultes

Abstract

En este artículo se realiza una discusión detallada de las principales metodologías empleadas para el estudio de fenómenos marginales y ocultos, como son el muestreo en bola de nieve, el método de estimación de captura-recaptura y el método para estimar el número de consumidores de cocaína en una gran ciudad. Este nuevo método se basa en dos selecciones aleatorias y puede estimar la ocultación del fenómeno.

This paper reviws some of the major methodologies applied to study rare and elusive populations as snowball sampling, capture-recapture technique and nominative methods. We also propose a new method to estimate the size of cocain-using population in a big city. This new method use two random selections and is capable to estimate even the size of the occultation of the drug-using phenomenon.

A interior point algorithm for linear programming is developed. At every step of the method a interior point to the polytope is available. With center at this point, an ellipsoid wholly included in the polytope is considered. The optimization of the linear objetive over the ellipsoid is carried out through the solution of a least squares type system of equations. The resulting interior point is adopted for the next iteration. As a result of this work, two different methods have been proposed for solving the least square problem, each showing a considerable decrease in the number of operations required in every step. Lastly, a computer application was developed for the method, in the Fortran 77, for an IBM PC AT, to carry out an experimental study of the behaviour of the algorithm in comparison with the Simplex Method.