Relatório de pesquisa 13/08

Diagnostics analysis for skew normal regression models, C. S. Ferreira, H. Bolfarino, and F. V. Labra, submitted 06/08

The skew normal model (Azzalini, 1985) is being used successfully in various statistical applications. The main purpose of this paper is to consider local influence analysis, which is a well-recognized  important step of data analysis. In this paper local influence measures are developed via Zhu and Lee's (2001) approach, that is closely related to the EM-algorithm. The diagnostic measures derived under this approach are invariant under reparameterization. Some useful perturbation schemes are discussed. The Cook's approach is also calculated and compared with Zhu and Lee's approach. Results that are obtained from analysis of a real data example are presented to illustrate the developed methodologies.
Primary 20M20, 93B29; Secondary 22E46.

Skew-normal distribution; Cook's approach; Zhu and Lee's approach, EM algorithm

Copy of the file:

rp13-08.pdf (PDF)

rp13-08.pdf.gz (gzipped PDF)

June26, 2008


Volta ao indíce de Relatórios de Pesquisa