Accession Number : ADA290694

Title :   Optimal Universal Coding and Density Estimation.

Descriptive Note : Final rept. 1 Jun 91-30 Sep 94,

Corporate Author : WISCONSIN UNIV-MADISON DEPT OF STATISTICS

Personal Author(s) : Yu, Bin

PDF Url : ADA290694

Report Date : 28 NOV 1994

Pagination or Media Count : 5

Abstract : Research progress has been made in the areas of empirical processes for mixing sequences, information theory, minimax estimation theory in source coding and nonparametric statistics, and Markov chain Monte Carlo (MCMC) methods. Rates of convergence and Central Limit Theorems results have been obtained for empirical processes of dependent data, and they are very useful for studying statistical models with dependence structure. On the important MCMC convergence diagnostic problem, regeneration points have been introduced into the Markov chain using the split-chain technique; so has been a global approach based on the the estimated L1 error and the Cusum path plot. Making connections between information theory and statistics, we obtained an information-theoretic result on the rate of convergence of a D-semifaithful code, and we also introduced non-parametric minimax lower bound techniques into bounding from below the redundancy in source coding. (AN)

Descriptors :   *MONTE CARLO METHOD, *MARKOV PROCESSES, MATHEMATICAL MODELS, COMPUTATIONS, PROBABILITY DENSITY FUNCTIONS, ESTIMATES, NONPARAMETRIC STATISTICS, CODING, CONVERGENCE, BAYES THEOREM, INFORMATION THEORY, SEQUENCES(MATHEMATICS).

Subject Categories : Statistics and Probability

Distribution Statement : APPROVED FOR PUBLIC RELEASE