A.C. Davison
Department of Mathematics, Swiss Federal Institute of Technology, Lausanne 1015, Switzerland
Bootstrap methods are simulation approaches to inference for statistical problems where large-sample approximations are unreliable, where parametric assumptions cannot be
trusted, or where the procedure to be applied is complicated or poorly-understood. The essential idea is repetition of the original procedure on bootstrap datasets, obtained by mimicking the sampling plan that led to the original data. This bootstrap output is then combined in such a way as to provide insight into the original problem, for example in the form of confidence intervals, prediction regions, or significance levels for tests.
The first part of this talk will describe bootstrap methods in general. I’ll then turn to regression problems, and finally talk about particular problems that arise in the context of population models.
Reference: PAGE 8 (1999) Abstr 75 [www.page-meeting.org/?abstract=75]
Poster: oral presentation