Séminaire de Probabilités et Statistique :

Le 22 septembre 2008 à 10:30 - UM2 - Bât 09 - Salle TD31

Présentée par Jhun Myoungshic - Department of Statistics, Korea University

"Bootstrap Methods for Categorical Data Analysis"

Since the bootstrap method was proposed by Efron(1979), it has been shown that the method can be used for the approximation of sampling distribution in many statistical problems. Theoretical properties of the method for multinomial problems were discussed by Efron(1982), Woodroofe & Jhun(1989) etc. We applied the bootstrap method for two specific problems in categorical data analysis, which are ‘simultaneous confidence regions for multinomial populations’ and ‘tests for independence in ordinal contingency tables’. Advantages of the bootstrap method are demonstratedby using simulation studies and real examples. Most of the material is based on the previous results including Woodroofe & Jhun(1989), Jhun and Jeong (2000), Jeong, Jhun & Kim(2005 ).