• Asian-Australasian Journal of Animal Sciences
• /
• v.14 no.8
• /
• pp.1051-1056
• /
• 2001

Data for teat number for Landrace (L), Yorkshire (Y), crossbred of Landrace and Yorkshire (LY), and crossbred of Landrace, Yorkshire and Chinese indigenous Min Pig (LYM) were analyzed using Gibbs sampling. In Bayesian inference, flat priors and some informative priors were used to examine their influence on posterior estimates. The posterior mean estimates of heritabilities with flat priors were $0.661{\pm}0.035$ for L, $0.540{\pm}0.072$ for Y, $0.789{\pm}0.074$ for LY, and $0.577{\pm}0.058$ for LYM, and they did not differ (p>0.05) from their corresponding estimates of REML. When inverse Gamma densities for variance components were used as priors with the shape parameter of 4, the posterior estimates were still corresponding (p>0.05) to REML estimates and mean estimates using Gibbs sampling with flat priors. However, when the inverse Gamma densities with the shape parameter of 10 were utilized, some posterior estimates differed (p<0.10) from REML estimates and/or from other Gibbs mean estimates. The use of moderate degree of belief was influential to the posterior estimates, especially for Y and for LY where data sizes were small. When the data size is small, REML estimates of variance components have unknown distributions. On the other hand, Bayesian approach gives exact posterior densities of variance components. However, when the data size is small and prior knowledge is lacked, researchers should be careful with even moderate priors.

• Journal of Animal Science and Technology
• /
• v.46 no.5
• /
• pp.719-728
• /
• 2004

The aims of this study were to estimate genetic parameters for carcass traits on Hanwoo(Korean Native Cattle) and to compare two different statistical algorithms for estimating genetic parameters. Data obtained from 1526 steers at Hanwoo Improvement Center and Hanwoo Improvement Complex Area from 1996 to 2001 were used for the analyses. The carcass traits considered in these studies were carcass weight, dressing percent, eye muscle area, backfat thickness, and marbling score. Estimated genetic parameters using EM-REML algorithm were compared to those by Bayesian inference via Gibbs Sampling to find out statistical properties. The estimated heritabilities of carcass traits by REML method were 0.28, 0.25, 0.35, 0.39 and 0.51, respectively and those by Gibbs Sampling method were 0.29, 0.25, 0.40, 0.42 and 0.54, respectively. This estimates were not significantly different, even though the estimated heritabilities by Gibbs Sampling method were higher than ones by REML method. Since the estimated statistics by REML method and Gibbs Sampling method were not significantly different in this study, it is inferred that both mothods could be efficiently applied for the analysis of carcass traits of cattle. However, further studies are demanded to define an optimal statistical method for handling large scale performance data.

• Communications for Statistical Applications and Methods
• /
• v.24 no.3
• /
• pp.291-301
• /
• 2017

It is well known in a small sample that the maximum likelihood (ML) approach for variance components in the general linear model yields estimates that are biased downward. The ML estimate of residual variance tends to be downwardly biased. The underestimation of residual variance, which has implications for the estimation of marginal effects and asymptotic standard error of estimates, seems to be more serious in some limited dependent variable models, as shown by some researchers. An alternative frequentist's approach may be restricted or residual maximum likelihood (REML), which accounts for the loss in degrees of freedom and gives an unbiased estimate of residual variance. In this situation, the REML estimator is derived in a censored regression model. A small sample the REML is shown to provide proper inference on regression coefficients.

• Asian-Australasian Journal of Animal Sciences
• /
• v.16 no.5
• /
• pp.639-643
• /
• 2003

The data on 283 Sahiwal cows, sired by 16 bulls, maintained at Cattle Breeding Farm of Nagpur Veterinary College and Dairy Farm of Agricultural College, Nagpur, were considered for the estimation of genetic parameters. Variance and covariance estimates of first lactation traits were obtained using restricted maximum likelihood technique (REML). When first lactation milk yield (FLMY), first lactation length (FLL) and average daily yield (ADY) traits were considered for REML analysis, the heritabilities were $0.184{\pm}0.146$, $0.132{\pm}0.131$ and $0.141{\pm}0.133$, respectively. While, genetic and phenotypic correlations between them were medium to high except phenotypic correlations between FLL and ADY (-0.025). REML procedure considering FLMY, age at first calving (AFC) and first service period (FSP) combination exhibits heritabilities as $0.274{\pm}0.173$, $0.506{\pm}0.233$ and $0.274{\pm}0.172$, respectively. Genetic correlations were $-0.120{\pm}0.376$, $0.225{\pm}0.423$ and $0.365{\pm}0.331$ between FLMY and AFC, FLMY and FSP, AFC and FSP, respectively. Phenotypic correlations were 0.057, 0.289 and 0.123, respectively. Considering all five traits REML combination heritabilities estimated were $0.238{\pm}0.162$, $0.160{\pm}0.139$, $0.136{\pm}0.132$, $0.409{\pm}0.209$ and $0.259{\pm}0.168$ for FLMY, FLL, ADY, AFC and FSP, respectively. The genetic correlations were positive except FLMY and AFC. The phenotypic correlations were also positive except FLL and ADY, ADY and FSP. Almost all estimates were associated with high standard error.

• The Korean Journal of Applied Statistics
• /
• v.15 no.2
• /
• pp.337-354
• /
• 2002

This paper derives three estimation methods for the between group variance component for serially correlated random model. To compare their estimation capability, three designs having different degree of unbalancedness are considered. The so-called empirical quantile dispersion graphs(EQDGs) used to compare estimation methods as well as designs. The proposed conditional ANOVA estimation is robust for design unbalancedness, however, ML estimation is preferred to the conditional AOVA and REML estimation regardless of design unbalancedness and correlation coefficient.

• Journal of Animal Science and Technology
• /
• v.44 no.4
• /
• pp.383-390
• /
• 2002

The data were consisted of 1,262 records for carcass traits observed at Hanwoo steers from 1998 to 2001 at Namwon and Deakwanryung branch of National Livestock Research Institute, Rural Development Administration. Pedigrees of young bulls were traced back to search for magnifying inbreeding. Genetic parameters for carcass traits with Gibbs sampling in a threshold animal model were compared to estimates with REML algorithm in linear model. As the results, most of bulls were not inbred and sire pedigree group was non-inbred population. However, most of the bulls fell in some relationship with each other. Heritability estimates as fully posterior means by Gibbs samplers in threshold model were higher than those by REML in linear model. Furthermore, these estimates in threshold model using GS showed higher estimates than estimates using tested young bulls in previous study and same model. Heritability estimate by GS for marbling score was 0.74 and genetic correlation estimate between marbling score and body weight at slaughter was –0.44. Further study for correlation of breeding values between REML algorithm in linear model and Gibbs sampling algorithm in threshold model was needed.

• The Korean Journal of Applied Statistics
• /
• v.13 no.1
• /
• pp.55-70
• /
• 2000

본 논문에서는 내포오차성분을 가지는 패널회귀모형에서 회귀계수에 대하여 다양한 추정량들을 유도하고, 추정량들의 효율성을 모의실험을 통하여 평균제곱오차의 기준에서 비교하였다. 모의실험 결과, 제안된 FGLS 추정량들은 GLS추정량과 효율성에서 서로 큰 차이를 보이지 않았으며, 계산상 더욱 복잡한 ML, REML 추정량 및 MIVQUE와 거의 비슷한 효율성을 보여주었다.

• The Korean Journal of Applied Statistics
• /
• v.6 no.2
• /
• pp.355-370
• /
• 1993

The estimations of variance components for the unbalanced one way random effects model when the underlying distributions are not necessarily normal are considered. ANOVA, REML, ML, MIVQUE, and MINQE estimators are compared with respect to their mean squared errors and biases through a simulation study. Explicit, computable expressions with no matrix inversion necessary are given for these estimators. An efficient rule to provide a prior guess of MINQE is given. Our results indicate that the efficiency of MINQE is excellent for arbitrary underlying distribution in the sense of MSE even in the presence of nontrivial bias. Also, MINQE is a worthwhile improvement over other estimators when kurtosis of underlying distributions become large 1.

• Asian-Australasian Journal of Animal Sciences
• /
• v.13 no.3
• /
• pp.413-422
• /
• 2000

In the class of models which include random effects, the variance component estimates are important to obtain accurate predictors and estimators. Variance component estimation is straightforward for balanced data but not for unbalanced data. Since orthogonality among factors is absent in unbalanced data, various methods for variance component estimation are available. REML estimation is the most widely used method in animal breeding because of its attractive statistical properties. Recently, Bayesian approach became feasible through Markov Chain Monte Carlo methods with increasingly powerful computers. Furthermore, advances in variance component estimation with complicated models such as generalized linear mixed models enabled animal breeders to analyze non-normal data.

• Journal of the Korean Data and Information Science Society
• /
• v.12 no.1
• /
• pp.19-25
• /
• 2001

Random effects models which describe the dependence via random effects in various correlated data have recently received considerable attention in the biomedical literature. They include mixed linear models (MLMs), generatized linear mixed models (GLMMS) and hierarchical generalized linear models (HGLMs). For the inference Lee and Nelder (2000) proposed the first-and second-order REML (restricted maximum likelihood) methods based on hierarchical-likelihood of tee and Welder (1996). In this paper, for Poisson-gamma HGLMs the new methods are theoretically compared with marginal likelihood methods and both methods are illustrated by two practical examples.