As far as the methodology is concerned, the heterogeneity has been dealt with mostly from a descriptive point of view, i.e. with the objective of determining those indicators capable of providing appropriate measurements of the degree of heterogeneity of a distribution. The object of this work is to establish an inferential procedure which allows for the solution of the above mentioned issues, whose goal is to compare the heterogeneity of two populations on the basis of sampling data. The method we propose consists of determining suitable test statistics and a general methodology based on the ordering of probabilities of various categories and on a nonparametric test, that, for the one-sided test, is similar to the one proposed by Pesarin (2001) for issues of stochastic dominance. It is important to note that the class probabilities of two distributions are unknown parameters and therefore the ordering of the probabilities can only be assessed on the basis of sampling data. The proposed solutions are therefore approximate.
One-sided and two-sided nonparametric tests for heterogeneity comparisons
BONNINI, Stefano;
2008
Abstract
As far as the methodology is concerned, the heterogeneity has been dealt with mostly from a descriptive point of view, i.e. with the objective of determining those indicators capable of providing appropriate measurements of the degree of heterogeneity of a distribution. The object of this work is to establish an inferential procedure which allows for the solution of the above mentioned issues, whose goal is to compare the heterogeneity of two populations on the basis of sampling data. The method we propose consists of determining suitable test statistics and a general methodology based on the ordering of probabilities of various categories and on a nonparametric test, that, for the one-sided test, is similar to the one proposed by Pesarin (2001) for issues of stochastic dominance. It is important to note that the class probabilities of two distributions are unknown parameters and therefore the ordering of the probabilities can only be assessed on the basis of sampling data. The proposed solutions are therefore approximate.I documenti in SFERA sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
