Information methods for model selection in linear mixed effects models with application to HCV data

In this paper, we derive a small sample Akaike information criterion, based on the maximized loglikelihood, and a small sample information criterion based on the maximized restricted loglikelihood in the linear mixed effects model when the covariance matrix of the random effects is known. Small sample corrected information criteria are proposed for a special case of linear mixed effects models, the balanced random-coefficient model, without assuming the random coefficients covariance matrix to be known. A simulation study comparing the derived criteria and several others for model selection in the linear mixed effects models is presented. We illustrate the behavior of the studied information criteria on real data from a study of subjects coinfected with HIV and Hepatitis C virus. Robustness of the criteria, in terms of the error distributed as a mixture of normal distributions, is also studied. Special attention is given to the behavior of the conditional AIC by Vaida and Blanchard (2005). Among the studied criteria, GIC performs best, while cAIC exhibits poor performance. Because of its inferior performance, as demonstrated in this work, we do not recommend its use for model selection in linear mixed effects models.     

Optimal Design of Blocked and Split-Plot Experiments for Fixed Effects and Variance Component Estimation

In general, modeling data from blocked and split-plot response surface experiments requires the use of generalized least squares and the estimation of two variance components. The literature on the optimal design of blocked and split-plot response surface experiments, however, focuses entirely on the precise estimation of the fixed factor effects and completely ignores the necessity to estimate the variance components as well. To overcome this problem, we propose a new Bayesian optimal design criterion which focuses on both the variance components and the fixed effects. A novel feature of the criterion is that it incorporates prior information about the variance components through log-normal or beta prior distributions. The resulting designs allow for a more powerful statistical inference than traditional optimal designs. In our algorithm for generating optimal blocked and split-plot designs, we implement efficient quadrature approaches for the numerical approximation of the new optimal design criterion. Supplementary materials for this article are available online.     

一个基于Web Service的异构信息集成方案

通过分析传统异构信息系统所存在的问题,提出一种基于Web Service技术的异构信息集成方案,建立统一的信息封装协议--REML.利用Web Service实现异构信息系统的重新利用,使得已有信息系统不必经过重新开发,就能被用户所访问,实现异构的信息资源集成共事.     

Estimation in mixed models via layer triangular transformation

A new alternative for estimating parameters in some mixed linear models is presented and illustrated. The approach is based on the same layer triangular transformation of the model matrix as proposed earlier by Öfversten (Biometrics 49 (1993)) for the derivation of some exact tests for variance components. Using the same preliminary transformation for both testing and estimation gives us an integrated set of procedures for the full analysis of some widely used mixed linear models. The proposed procedures are computationally stable and space saving.     

传感器实时信息共享的一种实现方案*

针对企业中会存在的各类传感器所产生的关于自然现象的一些实时信息,如果能够得到及时的分析处理,并作出相应的决策,将会极大地提高企业的运作效率.提出了一种在信息网格中集成传感器的方法,并以智能数字磁场计HMR2300为例说明其实现过程.实践表明,该方法能够更方便、更快捷地将传感器集成在信息网格中,实现实时信息共享.     

Extending the long-term survivor mixture model with random effects for clustered survival data

To provide a class of hazard functions in analyzing survival data, the power family of transformations has been proposed in the literature. Our work in this paper considers the existence of cured patients and random effects due to clustering of survival data in a long-term survivor model setting. A power family of transformations is assumed for the relative risk in the hazard function component. Such an extension allows us to flexibly base the inferences on various hazard function assumptions, particularly taking exponential and linear relative risk as two special cases. The parameter governing the power transformation could be determined by means of a modified Akaike information criterion (AIC). Applications to two sets of survival data illustrate the use of the proposed long-term survivor mixture model. A simulation study is carried out to examine the performance of the estimators under the proposed numerical estimation scheme.     

Joint Estimation of Location,Dispersion, and Random Effects in Robust Design

This article presents a methodology for simultaneously modeling three components of a general mixed-model approach to robust design—location (fixed) effects, dispersion effects, and random effects. Control, noise, and random factors can be accommodated with this approach. Parameters associated with all three are estimated jointly using residual maximum likelihood and assuming normality. Simulated and real datasets illustrate the key concepts and advantages over previously proposed approaches.     

Mixed model regression mapping for QTL detection in experimental crosses

This paper presents mixed model regression mapping (MMRM) as a method for mapping quantitative trait loci (QTL) in backcross and F2 data arising from crosses of inbred lines. It is related to interval mapping, composite interval mapping and other regression approaches but differs in that it tests for QTL presence in each linkage group before conditionally modeling QTL location.The three key ideas presented are to promote use of a Likelihood Ratio type of test for the presence of QTL in linkage groups before searching for QTL as a method of controlling false discovery rate, to present an alternative QTL profile to the LOD score for identifying the possible location of a QTL, and to promote the use of a local smoother to identify turning points in a profile based on evaluation at marker points rather than directly predicting all intermediate points.MMRM requires fitting a short series of models to locate and then evaluate putative QTL. Assuming marker covariates are allocated to linkage groups, MMRM first fits all the markers as independent random effects with common variance within the linkage groups. If there is no significant variance component associated with a linkage group, there is no evidence for a QTL associated with that group. Otherwise a QTL profile is predicted as a weighted sum of the marker BLUPs from which to postulate the most likely position of the QTL. A putative QTL covariate for that position is then calculated from flanking markers and added to the model. If this does not explain all the marker variance, the model is refined.Since MMRM is based on a linear mixed model, the model is easily extended to include extraneous sources of variation such as spatial variation in field experiments, to handle multiple QTL and to test for genotype by environment interactions. It is expounded using two simple examples analysed in the ASReml linear models software. Two simulation studies show that MMRM identifies QTL as reliably as but more directly than other common methods.     

Changes in genetic parameters for milk yield and heat tolerance in the Thai Holstein crossbred dairy population under different heat stress levels and over time

The objectives of this study were to investigate changes in genetic parameters for milk yield (MY) and heat tolerance of the crossbred Thai Holstein Friesian population under different heat stress levels over time, and to investigate the threshold point of heat stress manifestation on milk production. Genetic parameters were estimated using single-step genomic REML (ssGREML) and traditional REML models. Data included 58,965 test-day MY records from 1999 to 2008 (old data) and 105,485 test-day MY records from 2009 to 2018 (recent data) from the first parity of 24,520 cows. The pedigree included 55,168 animals, of which 882 animals had genotypes. Variance components were estimated with the REMLF90 program using a repeatability model with random regressions on a function of temperature-humidity index (THI) for additive genetic and permanent environmental effects. Fixed effects included farm-calving season combination, breed group-months in milk combination, and age at first calving. Random effects included additive genetic (intercept and slope) effects, permanent environmental (intercept and slope) effects, and herd-month-year of test. The phenotypic mean for MY was 13.33 ± 4.39 kg/d in the old data, and 14.48 ± 4.40 kg/d in the recent data. Estimates over different THI levels for the intercept additive genetic variance using old data ranged from 2.61 to 2.77 and from 5.02 to 5.38 using recent data with the REML method. In ssGREML analyses (performed with recent data only) the estimates for the intercept additive genetic variance ranged from 4.71 to 5.05. Estimates for the slope additive genetic variance were close to zero in all cases, with the largest values (0.024–0.030) at the most extreme THI value (80). Using REML, the covariance between the intercept and the slope additive genetic effects (THI from 72 to 80) ranged from −0.001 to 0.019 with old data and from 0.027 to 0.060 with recent data. The same covariance ranged from 0.026 to 0.057 in ssGREML analyses. The covariance between the intercept and the slope permanent environmental effects ranged from −0.42 to −0.67 for all data and THI levels. Across THI levels, the genetic correlation between MY and heat tolerance varied from −0.06 to 0.13 with old data, from 0.16 to 0.30 with recent data in REML analyses, and from 0.15 to 0.30 in ssGREML analyses, suggesting that in the current population the top animals for MY are more resistant to heat stress. This was expected, because of the introduction of Bos indicus genes in the last years. Heritability estimates for MY ranged from 0.19 to 0.21 (old data) and from 0.33 to 0.40 (recent data) for REML analyses. Heritability estimates for MY using ssGREML ranged from 0.31 to 0.38. A decline in MY was found when the animals' breed composition had more than 97.3% of Holstein genetics, and it was greatest at THI 80. The heritability and genetic correlations observed in this study show that selection for MY is possible without a negative correlated response for heat tolerance. Although the inclusion of genomic information is expected to increase the accuracy of selection, more genotypes must be collected for successful application. Future research should address other production and fitness traits within the Thai Holstein population.     

用层三角变换作混合模型中的参数估计

在混合线性模型中估计方差分量,最广泛应用的方法是极大似然(ML)法和约束极大似然(REML)法.从数理统计理论的高度来看,这些方法相当完美,但从数值计算的角度来看,却比较繁琐.为克服计算上的困难,在Oefversten[1]关于推导检验方差分量变换工作的基础上,我们提出了一种新的所谓“层三角变换法”。