A stage‐wise approach for the analysis of multi‐environment trials

Plant breeders and variety testing agencies routinely test candidate genotypes (crop varieties, lines, test hybrids) in multiple environments. Such multi‐environment trials can be efficiently analysed by mixed models. A single‐stage analysis models the entire observed data at the level of individual plots. This kind of analysis is usually considered as the gold standard. In practice, however, it is more convenient to use a two‐stage approach, in which experiments are first analysed per environment, yielding adjusted means per genotype, which are then summarised across environments in the second stage. Stage‐wise approaches suggested so far are approximate in that they cannot fully reproduce a single‐stage analysis, except in very simple cases, because the variance–covariance matrix of adjusted means from individual environments needs to be approximated by a diagonal matrix. This paper proposes a fully efficient stage‐wise method, which carries forward the full variance–covariance matrix of adjusted means from the individual environments to the analysis across the series of trials. Provided the variance components are known, this method can fully reproduce the results of a single‐stage analysis. Computations are made efficient by a diagonalisation of the residual variance–covariance matrix, which necessitates a corresponding linear transformation of both the first‐stage estimates (e.g. adjusted means and regression slopes for plot covariates) and the corresponding design matrices for fixed and random effects. We also exemplify the extension of the general approach to a three‐stage analysis. The method is illustrated using two datasets, one real and the other simulated. The proposed approach has close connections with meta‐analysis, where environments correspond to centres and genotypes to medical treatments. We therefore compare our theoretical results with recently published results from a meta‐analysis.     

A Method for Analyzing Nonlinear Models When the Data are from a Split-plot or Repeated Measures Design

Statistical methods are described which can be used to compare treatments where the response model is nonlinear and the experimental design includes split-plots or repeated measures. The nonlinear analysis of covariance is described for a two-way treatment structure in a split-plot design structure, the usual split-plot experimental design. Evaluating heat tolerance in common beans as a function of temperature is used as an example to demonstrate the methodology.     

Dose Response Analysis Using Robust Covariance Estimation

A dose response analysis is robustified by estimating the asymptotic covariance of the fitted model parameters by the approximate information sandwich (a sandwich statistic) under a heterogeneous variance. The robust method is described by using a nonlinear four‐parameter regression model. The usual, robust, bootstrap, and jackknife estimates of the asymptotic variance are examined for the bioassay data. Under the response of a normal distribution with changing variances over the dose levels, the performance of the usual and robust variances is investigated by Monte Carlo study. It confirms the robustness of the sandwich estimate and shows the non‐accuracy of the usual asymptotic variance estimates of fitted model parameters under the different forms of nonconstant variance structures.     

Estimators of repeatability

Summary Analysis of variance and principal components methods have been suggested for estimating repeatability. In this study, six estimation procedures are compared: ANOVA, principal components based on the sample covariance matrix and also on the sample correlation matrix, a related multivariate method (structural analysis) based on the sample covariance matrix and also on the sample correlation matrix, and maximum likelihood estimation. A simulation study indicates that when the standard linear model assumptions are met, the estimators are quite similar except when the repeatability is small. Overall, maximum likelihood appears the preferred method. If the assumption of equal variance is relaxed, the methods based on the sample correlation matrix perform better although others are surprisingly robust. The structural analysis method (with sample correlation matrix) appears to be best.Paper number 776 from the Department of Meat and Animal Science, University of Wisconsin-Madison.     

Modelling balanced longitudinal data: maximum likelihood estimation and analysis of variance.

For linear models, assuming a within-experimental-units covariance structure that incorporates errors of measurement, serial correlation, and variation between units, results on explicit estimation of regression parameters are used to simplify maximum likelihood estimation of covariance parameters. The use of an analysis of variance table as a simpler alternative to likelihood inference is illustrated with two examples.     

Maximum Likelihood based comparison of the specific growth rates for P. aeruginosa and four mutator strains

The specific growth rate for P. aeruginosa and four mutator strains mutT, mutY, mutM and mutY-mutM is estimated by a suggested Maximum Likelihood, ML, method which takes the autocorrelation of the observation into account. For each bacteria strain, six wells of optical density, OD, measurements are used for parameter estimation. The data is log-transformed such that a linear model can be applied. The transformation changes the variance structure, and hence an OD-dependent variance is implemented in the model. The autocorrelation in the data is demonstrated, and a correlation model with an exponentially decaying function of the time between observations is suggested. A model with a full covariance structure containing OD-dependent variance and an autocorrelation structure is compared to a model with variance only and with no variance or correlation implemented. It is shown that the model that best describes data is a model taking into account the full covariance structure. An inference study is made in order to determine whether the growth rate of the five bacteria strains is the same. After applying a likelihood-ratio test to models with a full covariance structure, it is concluded that the specific growth rate is the same for all bacteria strains. This study highlights the importance of carrying out an explorative examination of residuals in order to make a correct parametrization of a model including the covariance structure. The ML method is shown to be a strong tool as it enables estimation of covariance parameters along with the other model parameters and it makes way for strong statistical tools for inference studies.     

The Co-Occurrence of Species and the Co-Diversity of Sites in Neutral Models of Biodiversity

Patterns of co-occurrence of species are widely used to assess the fit of ecological neutral models to empirical patterns. The mathematically equivalent patterns of co-diversity of sites, in contrast, have been considered only indirectly and analyses normally are focused on the spatial distribution of species richness, rather than on the patterns of species sharing. Here we use two analytical tools (range-diversity plots and rank plots) to assess the predictions of simple neutral models in relation to patterns of co-occurrence and co-diversity. Whereas a fully stochastic null model predicts zero average among species and among sites, neutral models generate systems with low levels of covariance among species and high levels of positive covariance among sites. These patterns vary with different combinations of dispersal and speciation rates, but are always linked to the shape, symmetry, and spread of the range-size and species-richness frequency distributions. Non-homogeneous patterns of diversity and distribution arise in neutral models because of the spatial arrangement of sites and their concomitant similarity, which is reflected also in the spread of the range-size frequency distribution. The nearly null covariance among species, in contrast, implies low variance in species richness of sites and very slim frequency distributions. In real world assemblages of Mexican volant and non-volant mammals, patterns of range-size and species-richness frequency distribution are similar to those generated by neutral models. However, when the comparison includes the covariance both for species (co-occurrence) and for sites (co-diversity), empirical patterns differ significantly from the predictions of neutral models. Because of the mathematical links between the covariance in the distribution of species and the variance of species-richness values and between the covariance in species sharing among sites and the variance of range-size values, a full understanding of patterns of diversity calls for the simultaneous analysis of co-occurrence and co-diversity.     

Analysis of covariance with pre-treatment measurements in randomized trials: comparison of equal and unequal slopes

In randomized trials, an analysis of covariance (ANCOVA) is often used to analyze post-treatment measurements with pre-treatment measurements as a covariate to compare two treatment groups. Random allocation guarantees only equal variances of pre-treatment measurements. We hence consider data with unequal covariances and variances of post-treatment measurements without assuming normality. Recently, we showed that the actual type I error rate of the usual ANCOVA assuming equal slopes and equal residual variances is asymptotically at a nominal level under equal sample sizes, and that of the ANCOVA with unequal variances is asymptotically at a nominal level, even under unequal sample sizes. In this paper, we investigated the asymptotic properties of the ANCOVA with unequal slopes for such data. The estimators of the treatment effect at the observed mean are identical between equal and unequal variance assumptions, and these are asymptotically normal estimators for the treatment effect at the true mean. However, the variances of these estimators based on standard formulas are biased, and the actual type I error rates are not at a nominal level, irrespective of variance assumptions. In equal sample sizes, the efficiency of the usual ANCOVA assuming equal slopes and equal variances is asymptotically the same as those of the ANCOVA with unequal slopes and higher than that of the ANCOVA with equal slopes and unequal variances. Therefore, the use of the usual ANCOVA is appropriate in equal sample sizes.     

Maximum Likelihood based comparison of the specific growth rates for P. aeruginosa and four mutator strains

The specific growth rate for P. aeruginosa and four mutator strains mutT, mutY, mutM and mutY–mutM is estimated by a suggested Maximum Likelihood, ML, method which takes the autocorrelation of the observation into account. For each bacteria strain, six wells of optical density, OD, measurements are used for parameter estimation. The data is log-transformed such that a linear model can be applied. The transformation changes the variance structure, and hence an OD-dependent variance is implemented in the model. The autocorrelation in the data is demonstrated, and a correlation model with an exponentially decaying function of the time between observations is suggested. A model with a full covariance structure containing OD-dependent variance and an autocorrelation structure is compared to a model with variance only and with no variance or correlation implemented. It is shown that the model that best describes data is a model taking into account the full covariance structure. An inference study is made in order to determine whether the growth rate of the five bacteria strains is the same. After applying a likelihood-ratio test to models with a full covariance structure, it is concluded that the specific growth rate is the same for all bacteria strains. This study highlights the importance of carrying out an explorative examination of residuals in order to make a correct parametrization of a model including the covariance structure. The ML method is shown to be a strong tool as it enables estimation of covariance parameters along with the other model parameters and it makes way for strong statistical tools for inference studies.     

人类群体遗传结构的协方差阵主成分分析方法

目的:探讨基因频率矩阵的中心化(或均值化)协方差阵主成分分析方法在人类群体遗传结构研究中的适用性和合理性。方法:从基因频率矩阵的结构特征入手,分析中心化、均值化协方差阵主成分分析与标准化相关阵主成分分析在特征根、特征向量以及降维效果等方面的差异,并通过实例比较不同方法在解释群体遗传结构特征上合理性。结果:中心化(或均值化)协方差阵的主成分不仅反映了基因变异程度的“方差信息量权”,而且反映了基因间相互影响程度的“相关信息量权”;标准化相关阵的主成分反映的仅是“相关信息量权”,不包括“方差信息量权”。通过比较中国26个汉族人群HLA-A基因座中心化协方差阵和标准化相关阵2种主成分分析结果,证实中心化协方差阵主成分分析方法在特征根与特征向量、保留主成分的个数和对主成分的群体遗传学解释的合理性等方面均优于标准化相关阵主成分分析方法。结论:在对群体遗传结构进行主成分分析时,应使用中心化(或均值化)变换消除基因频率矩阵中量级的影响,然后在用其协方差阵提取主成分。     

A modified pseudolikelihood approach for analysis of longitudinal data

We consider the analysis of longitudinal data when the covariance function is modeled by additional parameters to the mean parameters. In general, inconsistent estimators of the covariance (variance/correlation) parameters will be produced when the "working" correlation matrix is misspecified, which may result in great loss of efficiency of the mean parameter estimators (albeit the consistency is preserved). We consider using different "working" correlation models for the variance and the mean parameters. In particular, we find that an independence working model should be used for estimating the variance parameters to ensure their consistency in case the correlation structure is misspecified. The designated "working" correlation matrices should be used for estimating the mean and the correlation parameters to attain high efficiency for estimating the mean parameters. Simulation studies indicate that the proposed algorithm performs very well. We also applied different estimation procedures to a data set from a clinical trial for illustration.     

Genetic analysis of crossfostering data with sire and dam records

A method is presented for the analysis of data from crossfostering experiments in which parts of litters are reciprocally interchanged at birth. Observed variances and covariances of differently related individuals are expressed as functions of theoretical causal components of phenotypic variance (additive direct, dominance direct, additive maternal, dominance maternal, direct-maternal covariance, and environmental). Causal components are estimated by weighted least squares analysis of this system of equations, including a ridge-regression procedure to examine consequences of correlation between observed components. Ridge regression suggests that dominance direct genetic variance is generally underestimated, but that narrow-sense heritability estimates are reliable.     

Analysis of longitudinal data with unmeasured confounders.

Confounding in longitudinal or clustered data creates special problems and opportunities because the relationship between the confounder and covariate of interest may differ across and within individuals or clusters. A well-known example of such confounding in longitudinal data is the presence of cohort and period effects in models of aging in epidemiologic research. We first formulate a data-generating model with confounding and derive the distribution of the response variable unconditional on the confounder. We then examine the properties of the regression coefficient for some analytic approaches when the confounder is omitted from the fitted model. The expected value of the regression coefficient differs in across- and within-individual regression. In the multivariate case, within- and between-individual information is combined and weighted according to the assumed covariance structure. We assume compound symmetry in the fitted covariance matrix and derive the variance, bias, and mean squared error of the slope estimate as a function of the fitted within-individual correlation. We find that even in this simplest multivariate case, the trade-off between bias and variance depends on a large number of parameters. It is generally preferable to fit correlations somewhat above the true correlation to minimize the effect of between-individual confounders or cohort effects. Period effects can lead to situations where it is advantageous to fit correlations that are below the true correlation. The results highlight the trade-offs inherent in the choice of method for analysis of longitudinal data, and show that an appropriate choice can be made only after determining whether within- or between-individual confounding is the major concern.     

THE MIXED-MODEL ANALYSIS OF VARIANCE APPLIED TO QUANTITATIVE GENETICS: BIOLOGICAL MEANING OF THE PARAMETERS

The mixed-model factorial analysis of variance has been used in many recent studies in evolutionary quantitative genetics. Two competing formulations of the mixed-model ANOVA are commonly used, the “Scheffe” model and the “SAS” model; these models differ in both their assumptions and in the way in which variance components due to the main effect of random factors are defined. The biological meanings of the two variance component definitions have often been unappreciated, however. A full understanding of these meanings leads to the conclusion that the mixed-model ANOVA could have been used to much greater effect by many recent authors. The variance component due to the random main effect under the two-way SAS model is the covariance in true means associated with a level of the random factor (e.g., families) across levels of the fixed factor (e.g., environments). Therefore the SAS model has a natural application for estimating the genetic correlation between a character expressed in different environments and testing whether it differs from zero. The variance component due to the random main effect under the two-way Scheffe model is the variance in marginal means (i.e., means over levels of the fixed factor) among levels of the random factor. Therefore the Scheffe model has a natural application for estimating genetic variances and heritabilities in populations using a defined mixture of environments. Procedures and assumptions necessary for these applications of the models are discussed. While exact significance tests under the SAS model require balanced data and the assumptions that family effects are normally distributed with equal variances in the different environments, the model can be useful even when these conditions are not met (e.g., for providing an unbiased estimate of the across-environment genetic covariance). Contrary to statements in a recent paper, exact significance tests regarding the variance in marginal means as well as unbiased estimates can be readily obtained from unbalanced designs with no restrictive assumptions about the distributions or variance-covariance structure of family effects.     

Generalized variance component models for clustered categorical response variables

A generalized variance component model is proposed for the analysis of a categorical response variable with extra-multinomial variation. Categorical data obtained from research designs such as randomized multicenter clinical trials or complex sample surveys with clustering frequently exhibit extra-variation resulting from intracluster correlation. General correlation patterns are accounted for by utilizing a mixed-effects modelling approach, estimating the cluster variance components through the method of moments and modelling functions of the observed proportions through the use of estimating equations. A flexible set of assumptions characterizing the underlying covariance structure for the proportions can be accommodated. The importance of accounting for extra-variation when performing hypothesis tests is highlighted with an application to data from a multi-investigator clinical trial.     

流域环境要素空间尺度特征及其与水生态分区尺度的关系——以辽河流域为例

指标体系构建是流域水生态分区技术框架中的一项重要内容。目前提出的指标体系尚缺乏一定的科学理论支持,如所选指标的生态尺度与各级分区大小之间的对应关系。生态学研究认为,不同尺度上的生态过程和格局不同。以辽河流域为例,对影响水生态系统的几个大尺度环境要素(降水、地形和植被)的空间尺度特征进行分析,得出了各要素空间变异最为显著的尺度。其中,降水的空间尺度约为75 km,地形要素和指标均大致存在16 km、32 km、64 km和128 km多个景观特征尺度。在界定水生态一、二级分区范围基础上,讨论了各环境要素作为水生态一、二级分区指标的适用性,以期为辽河流域水生态分区指标选取提供一定科学依据,同时希望能为其他流域提供一定参考。     

A comparison of likelihood ratio tests and Rao's score test for three separable covariance matrix structures

The problem of testing the separability of a covariance matrix against an unstructured variance‐covariance matrix is studied in the context of multivariate repeated measures data using Rao's score test (RST). The RST statistic is developed with the first component of the separable structure as a first‐order autoregressive (AR(1)) correlation matrix or an unstructured (UN) covariance matrix under the assumption of multivariate normality. It is shown that the distribution of the RST statistic under the null hypothesis of any separability does not depend on the true values of the mean or the unstructured components of the separable structure. A significant advantage of the RST is that it can be performed for small samples, even smaller than the dimension of the data, where the likelihood ratio test (LRT) cannot be used, and it outperforms the standard LRT in a number of contexts. Monte Carlo simulations are then used to study the comparative behavior of the null distribution of the RST statistic, as well as that of the LRT statistic, in terms of sample size considerations, and for the estimation of the empirical percentiles. Our findings are compared with existing results where the first component of the separable structure is a compound symmetry (CS) correlation matrix. It is also shown by simulations that the empirical null distribution of the RST statistic converges faster than the empirical null distribution of the LRT statistic to the limiting χ² distribution. The tests are implemented on a real dataset from medical studies.     

A comparison of phenotypic variation and covariation patterns and the role of phylogeny, ecology, and ontogeny during cranial evolution of new world monkeys

Similarity of genetic and phenotypic variation patterns among populations is important for making quantitative inferences about past evolutionary forces acting to differentiate populations and for evaluating the evolution of relationships among traits in response to new functional and developmental relationships. Here, phenotypic co variance and correlation structure is compared among Platyrrhine Neotropical primates. Comparisons range from among species within a genus to the superfamily level. Matrix correlation followed by Mantel's test and vector correlation among responses to random natural selection vectors (random skewers) were used to compare correlation and variance/covariance matrices of 39 skull traits. Sampling errors involved in matrix estimates were taken into account in comparisons using matrix repeatability to set upper limits for each pairwise comparison. Results indicate that covariance structure is not strictly constant but that the amount of variance pattern divergence observed among taxa is generally low and not associated with taxonomic distance. Specific instances of divergence are identified. There is no correlation between the amount of divergence in covariance patterns among the 16 genera and their phylogenetic distance derived from a conjoint analysis of four already published nuclear gene datasets. In contrast, there is a significant correlation between phylogenetic distance and morphological distance (Mahalanobis distance among genus centroids). This result indicates that while the phenotypic means were evolving during the last 30 millions years of New World monkey evolution, phenotypic covariance structures of Neotropical primate skulls have remained relatively consistent. Neotropical primates can be divided into four major groups based on their feeding habits (fruit-leaves, seed-fruits, insect-fruits, and gum-insect-fruits). Differences in phenotypic covariance structure are correlated with differences in feeding habits, indicating that to some extent changes in interrelationships among skull traits are associated with changes in feeding habits. Finally, common patterns and levels of morphological integration are found among Platyrrhine primates, suggesting that functional/developmental integration could be one major factor keeping covariance structure relatively stable during evolutionary diversification of South American monkeys.     

Ability of geometric morphometric methods to estimate a known covariance matrix

Landmark-based morphometric methods must estimate the amounts of translation, rotation, and scaling (or, nuisance) parameters to remove nonshape variation from a set of digitized figures. Errors in estimates of these nuisance variables will be reflected in the covariance structure of the coordinates, such as the residuals from a superimposition, or any linear combination of the coordinates, such as the partial warp and standard uniform scores. A simulation experiment was used to compare the ability of the generalized resistant fit (GRF) and a relative warp analysis (RWA) to estimate known covariance matrices with various correlations and variance structures. Random covariance matrices were perturbed so as to vary the magnitude of the average correlation among coordinates, the number of landmarks with excessive variance, and the magnitude of the excessive variance. The covariance structure was applied to random figures with between 6 and 20 landmarks. The results show the expected performance of GRF and RWA across a broad spectrum of conditions. The performance of both GRF and RWA depended most strongly on the number of landmarks. RWA performance decreased slightly when one or a few landmarks had excessive variance. GRF performance peaked when approximately 25% of the landmarks had excessive variance. In general, both RWA and GRF performed better at estimating the direction of the first principal axis of the covariance matrix than the structure of the entire covariance matrix. RWA tended to outperform GRF when > approximately 75% of the coordinates had excessive variance. When < 75% of the coordinates had excessive variance, the relative performance of RWA and GRF depended on the magnitude of the excessive variance; when the landmarks with excessive variance had standard deviations (sigma) > or = 4 sigma minimum, GRF regularly outperformed RWA.     

Developmental regulation of skull morphology II: ontogenetic dynamics of covariance

Canalization may play a critical role in molding patterns of integration when variability is regulated by the balance between processes that generate and remove variation. Under these conditions, the interaction among those processes may produce a dynamic structure of integration even when the level of variability is constant. To determine whether the constancy of variance in skull shape throughout most of postnatal growth results from a balance between processes generating and removing variation, we compare covariance structures from age to age in two rodent species, cotton rats (Sigmodon fulviventer) and house mice (Mus musculus domesticus). We assess the overall similarity of covariance matrices by the matrix correlation, and compare the structures of covariance matrices using common subspace analysis, a method related to common principal components (PCs) analysis but suited to cases in which variation is so nearly spherical that PCs are ambiguous. We find significant differences from age to age in covariance structure and the more effectively canalized ones tend to be least stable in covariance structure. We find no evidence that canalization gradually and preferentially removes deviations arising early in development as we might expect if canalization results from compensatory differential growth. Our results suggest that (co)variation patterns are continually restructured by processes that equilibrate variance, and thus that canalization plays a critical role in molding patterns of integration.