On the risk performance of extended sequence compound rules for classification between N(–1, 1) and N(1, 1)†
- 1 January 1978
- journal article
- research article
- Published by Taylor & Francis in Journal of Statistical Computation and Simulation
- Vol. 6 (3-4) , 265-280
- https://doi.org/10.1080/00949657808810195
Abstract
Since Robbins (1951) first introduced the compound decision problem, there has evolved a large literature on the subject for the most part dealing with the construction of compound rules whose excess risk over the simple envelope is no greater than zero in the limit as the number N of component problems goes to infinity. Such rules have compound risk which is asymptotically subminimax. Johns (1967) has introduced more stringent (extended) envelopes and has proposed extended compound rules whose risks achieve these envelopes in the limit. This paper reports some Monte Carlo results on the compound risk behavior of selected unextended and extended rules for moderate N values and certain parameter sequences for Robbins original example. The results show that the extended rules compare favorably with the minimax rule and the unextended rules for moderate N and parameter sequences exhibiting higher order empirical dependencies, for example, those generated by a Markov process.Keywords
This publication has 13 references indexed in Scilit:
- Asymptotic Solutions to the Two State Component Compound Decision Problem, Bayes Versus Diffuse Priors on ProportionsThe Annals of Statistics, 1976
- A smooth empirlcal bayes estimator for the mean of a multivariate normal distributionJournal of Statistical Computation and Simulation, 1974
- Optimal linear estimators: an empirical Bayes version with application to the binomial distributionBiometrika, 1971
- The performance of established and modified compound decision rulesBiometrika, 1971
- On an Extended Compound Decision ProblemThe Annals of Mathematical Statistics, 1969
- An empirical Bayes smoothing techniqueBiometrika, 1969
- The Sequential Compound Decision Problem with $m \times n$ Finite Loss MatrixThe Annals of Mathematical Statistics, 1966
- The Empirical Bayes Approach to Statistical Decision ProblemsThe Annals of Mathematical Statistics, 1964
- Asymptotic Solutions of the Sequential Compound Decision ProblemThe Annals of Mathematical Statistics, 1963
- Two Breakthroughs in the Theory of Statistical Decision MakingRevue de l'Institut International de Statistique / Review of the International Statistical Institute, 1962