Testing for Skew-Symmetric Models

Skew-symmetric (ss) models are semiparametric models for continuous random vectors. Starting with a symmetric probability density function (pdf), a ss family is generated by multiplying the symmetric pdf by a skewing function. A characterization of this skewing function let us propose a test for testing goodness of fit to a ss model. The test statistic is based on comparing two estimators of the population pdf: a consistent kernel-based estimator and another estimator which is built by assuming that the ss in the null hypothesis is true. The proposed test shares some properties with other tests based on comparing a kernel-based estimator of the population pdf to an estimator derived under the null hypothesis, like the ones in Hall (J. Multivar. Anal. 14, 1984, 1--16) and Fan (Econometric Theory, 10, 1994, 316--356), and other are new.Non UBCUnreviewedAuthor affiliation: University of SevillaFacult