Quantization of Models: Local Approach and Asymptotically Optimal Partitions
In this paper we review algorithmic aspects related to maximum-supportplane partitions. These partitions have been defined in Bock (1992) and analyzed in Pötzelberger and Strasser (2001). The local approach to inference leads to a certain subclass of partitions. They are obtained from quantizing the distribution of the score function. We propose a numerical method to compute these partitions B approximately, in the sense that they are asymptotically optimal for increasing sizes |B|. These findings are based on recent results on the asymptotic distribution of sets of prototypes.
KeywordsScore Function Relative Efficiency Quantization Error Optimal Partition Local Alternative
Unable to display preview. Download preview PDF.
- POTZELBERGER, K. (2002): Admissible unbiased quantizations: Distributions without linear components. To appear in: Math. Methods Statist.Google Scholar
- STEINER, G. (1999): Quantization and clustering with maximal information: Algorithms and numerical experiments, Ph.D. Thesis, Vienna University of Economics and Business Administration.Google Scholar
- STRASSER, H. (2000): Towards a statistical theory of optimal quantization. In W. Gaul, O. Opitz, M. Schader (eds.): Data Analysis: Scientific Modeling and Practical Application, Springer, Berlin Heidelberg, 369–383.Google Scholar
- ZADOR, P.L. (1964): Development and evaluation of procedures for quantizing Multivariate distributions, Ph.D. Thesis, Stanford University.Google Scholar