Convergence Property of the Estimate of Variance Ratio in the Tilde-Hat Procedure with Balanced Data

DOI Web Site 1 Citations 29 References

Bibliographic Information

Other Title
  • 釣合い型データに適用されたTilde-Hat法における分散比推定値の収束特性
  • Convergence Property of the Estimate of

Search this article

Abstract

The tilde-hat (TH) procedure is one of the pseudoexpectation (PE) methods for variance component estimation and is an approximation to the restricted maximum likelihood (REML) method. This procedure is sometimes used for estimating variance components and thereby genetic parameters using large data sets. Considering a balanced data-structure, in this paper, we examine the convergence property of the TH estimate of the variance ratio relative to the REML method using the expectation-maximization algorithm (EM-REML). A cross-classified mixed linear model containing one random effect except for the residual term is assumed. Under the current data-structure, the TH estimator is found to be identical to the REML estimator. However, the function representing the relationship between two successive estimates in the iteration is nonlinear in the case of EM-REML, while that in the TH estimation is shown to be linear. Using the functions derived, it is revealed that when normal convergence occurs with balanced data, the TH estimate always converges faster than the EM-REML estimate. Some discussion is given about the fact that the different PE methods applied to unbalanced data lead to different convergence patterns.

Journal

Citations (1)*help

See more

References(29)*help

See more

Keywords

Details 詳細情報について

Report a problem

Back to top