Asymptotic properties and numerical comparison of spacings-based power divergence statistics

0365979 - ÚTIA 2012 CZ eng A - Abstract
Boček, Pavel - Vajda, Igor - van der Meulen, Edward
Asymptotic properties and numerical comparison of spacings-based power divergence statistics.
Abstracts of Contributions to 6th International Workshop on Data – Algorithms – Decision Making. Praha: ÚTIA AV ČR, 2010 - (Ivánek, J.; Janžura, M.). s. 26-26
[International Workshop Jindřichův Hradec DAR. 2.12.2010, Jindřichův Hradec]
R&D Projects: GA MŠMT 1M0572
Institutional research plan: CEZ:AV0Z10750506
Keywords : divergence
Subject RIV: BB - Applied Statistics, Operational Research
http://library.utia.cas.cz/separaty/2010/SI/bocek-asymptotic properties and numerical comparison of spacings-based power divergence statistics.pdf

Measures of dissimilarity in the goodness of fit tests are defined as an appropriately scaled divergence between kvantizovanými hypothetical and empirical distributions. It is shown that hypothetical quantization generates statistical criterion clasická Pearson type, while the empirical quantization generates asymptotically equivalent to the classical criteria for incremental statistical criteria. Asymptotic properties are derived empirically generated criteria and proved quadratic optimality criteria.
Permanent Link: http://hdl.handle.net/11104/0201096