Facultad de Ciencias Económicas y de Administración, Universidad Nacional de Catamarca, 4700-Catamarca, San Fernando del Valle de, Argentina
Abstract:
In this work a device which changes the problem of mean estimation into that of proportion estimation is proposed. The device consists of perturbing the observations. The goal of the work is the construction of conservative confidence intervals for means. Three applications are given: (1) proportion estimation in the context of cluster random sampling, (2) differences of proportions of a multinomial population and (3) variance estimation.