Bias in genetic variance estimates due to spatial autocorrelation

Summary A central problem in the analysis of genetic field trials is the dichotomy of genetic and environmental effects because one cannot be defined without the other. Results from 768,000 simulated family trials in complete randomized block designs demonstrated a serious upward bias in estimates of family variance components from multi-unit plot designs when the phenotypic observations were compatible with a first-order autoregressive (AR1) process. The inflation of family variances and, thus, additive genetic variance and narrow sense individual heritabilities progressed exponentially with an increase in the nearest neighbor correlation () in the AR1 process. Significant differences in inflation rates persisted among various plot configurations. At = 0.2 the inflation of family variances reached 48–73%. Inflation rates were independent of the level of heritability. Modified Papadakis nearest neighbor (NN) adjustment procedures were tested for their ability to remove the bias in family variances. A NN-adjustment based on Mead's coefficient of interplant interaction and one derived from Bartlett's simultaneous autoregressive scheme removed up to 97% of the bias introduced by the phenotypic correlations. NN-adjusted estimates had slightly (5–8%) higher relative errors than did unadjusted estimates.     

Robust estimation of multivariate covariance components

In many settings, such as interlaboratory testing, small area estimation in sample surveys, and heritability studies, investigators are interested in estimating covariance components for multivariate measurements. However, the presence of outliers can seriously distort estimates obtained using standard procedures such as maximum likelihood. We propose a procedure based on M-estimation for robustly estimating multivariate covariance components in the presence of outliers; the procedure applies to balanced and unbalanced data. We present an algorithm for computing the robust estimates and examine the performance of the estimator through a simulation study. The estimator is used to find covariance components and identify outliers in a study of variability of egg length and breadth measurements of American coots.     

Spatial variance and spatial skewness: leading indicators of regime shifts in spatial ecological systems

Ecosystems can undergo large-scale changes in their states, known as catastrophic regime shifts, leading to substantial losses to services they provide to humans. These shifts occur rapidly and are difficult to predict. Several early warning signals of such transitions have recently been developed using simple models. These studies typically ignore spatial interactions, and the signal provided by these indicators may be ambiguous. We employ a simple model of collapse of vegetation in one and two spatial dimensions and show, using analytic and numerical studies, that increases in spatial variance and changes in spatial skewness occur as one approaches the threshold of vegetation collapse. We identify a novel feature, an increasing spatial variance in conjunction with a peaking of spatial skewness, as an unambiguous indicator of an impending regime shift. Once a signal has been detected, we show that a quick management action reducing the grazing activity is needed to prevent the collapse of vegetated state. Our results show that the difficulties in obtaining the accurate estimates of indicators arising due to lack of long temporal data can be alleviated when high-resolution spatially extended data are available. These results are shown to hold true independent of various details of model or different spatial dispersal kernels such as Gaussian or heavily fat tailed. This study suggests that spatial data and monitoring multiple indicators of regime shifts can play a key role in making reliable predictions on ecosystem stability and resilience. Electronic supplementary material The online version of this article (doi:) contains supplementary material, which is available to authorized users.     

Theoretical studies on the necessary number of components in mixtures

Summary Theoretical studies on the optimal numbers of components in mixtures (for example multiclonal varieties or mixtures of lines) have been performed according to phenotypic yield stability (measured by the parameter variance). For each component i, i = 1, 2,..., n, a parameter u_i with 0 u_i 1 has been introduced reflecting the different survival and yielding ability of the components. For the stochastic analysis the mean of each u_i is denoted by u₁ and its variance by _i²For the character total yield the phenotypic variance V can be explicitly expressed dependent on 1) the number n of components in the mixture, 2) the mean of the _i²3) the variance of the _i²4) the ratio and 5) the ratio _i²/² where denotes the mean of the u_i and _u²is the variance of the u_j. According to the dependence of the phenotypic stability on these factors some conclusions can be easily derived from this V-formula. Furthermore, two different approaches for a calculation of necessary or optimal numbers of components using the phenotypic variance V are discussed: A. Determination of optimal numbers in the sense that a continued increase of the number of components brings about no further significant effect according to stability. B. A reduction of b % of the number of components but nevertheless an unchanged stability can be realized by an increase of the mean of the u_i by 1% (with and _u²assumed to be unchanged). Numerical results on n (from A) and 1 (from B) are given. Computing the coefficient of variation v for the character total yield and solving for the number n of components one obtains an explicit expression for n dependent on v and the factors 2.-5. mentioned above. In the special case of equal variances, _i²= _o²for each i, the number n depends on v, x = (₀/)² and y = (_u/)². Detailed numerical results for n = n (v, x, y) are given. For x 1 and y 1 one obtains n = 9, 20 and 79 for v = 0.30, 0.20 and 0.10, respectively while for x 1 and arbitrary y-values the results are n = 11, 24 and 95.This publication is an extended version of a lecture given at the 1984-EUCARPIA meeting (Section Biometrics in Plant Breeding) in Stuttgart-Hohenheim (Federal Republic of Germany)     

On the interpretation of principal components analysis in ecological contexts

Summary Principal components analysis is well suited for many data analysis problems in ecology, particularly for data reduction and hypothesis generation; but the structure of PCA is poorly suited for indirect gradient analysis. Whatever the intended application of PCA, the user must exercise special care in selecting data transformations to prevent the analysis from being overwhelmed by the purely numerical effects in the variance structure of the data.I would like to thank R. H. Whittaker, H. G. Gauch, R. E. Moeller, and S. R. Searle for their guidance and assistance.     

Predicting mean and variance of all possible lines and hybrids from designs with partially inbred progenies

Two-factor mating designs at consecutive S_n and S_(n+1) levels (S₀ and S₁ S₁ and S₂, or F₂ and F₃) allow estimation of all components of the variation among homozygous lines and F₁ hybrids that can be derived from a given population. They also allow for the prediction of the mean of these lines and single-cross hybrids. Some tests for the presence of epistasis are possible at the levels of means and of variances. Such mating designs can be very useful for predicting the value of the best possible lines or the best possible F₁ hybrids when it is difficult to produce, at an experimental level for exploratory purposes, either lines or hybrids.     

The effect of genetic drift on the variance/covariance components generated by multilocus additive x additive epistatic systems

The effect of population bottlenecks on the components of the genetic variance/covariance generated by n neutral independent additive x additive loci has been studied theoretically. In its simplest version, this situation can be modelled by specifying the allele frequencies and homozygous effects at each locus, and an additional factor measuring the strength of the n-th order epistatic interaction. The variance/covariance components in an infinitely large panmictic population (ancestral components) were compared with their expected values at equilibrium over replicates randomly derived from the base population, after t bottlenecks of size N (derived components). Formulae were obtained giving the derived components (and the between-line variance) as functions of the ancestral ones (alternatively, in terms of allele frequencies and effects) and the corresponding inbreeding coefficient F(t). The n-th order derived component of the genetic variance/covariance is continuously eroded by inbreeding, but the remaining components may increase initially until a critical F(t) value is attained, which is inversely related to the order of the pertinent component, and subsequently decline to zero. These changes can be assigned to the between-line variances/covariances of gene substitution and epistatic effects induced by drift. Numerical examples indicate that: (1) the derived additive variance/covariance component will generally exceed its ancestral value unless epistasis is weak; (2) the derived epistatic variance/covariance components will generally exceed their ancestral values unless allele frequencies are extreme; (3) for systems showing equal ancestral additive and total non-additive variance/covariance components, those including a smaller number of epistatic loci may generate a larger excess in additive variance/covariance after bottlenecks than others involving a larger number of loci, provided that F(t) is low. Our results indicate that it is unlikely that the rate of evolution may be significantly accelerated after population bottlenecks, in spite of occasional increments of the derived additive variance over its ancestral value.     

Vegetation establishment in post-fire Adenostoma chaparral in relation to fine-scale pattern in fire intensity and soil nutrients

At fine spatial scales (0.1–10m), chaparral communities have been shown to be strongly influenced by canopy-gap patterns, leading to periodicities in vegetation at 4–5 m spatial scales. Fine-scale variations in fire behavior and post-fire erosion can lead to changes in the patterning of viable seeds and nutrients and may alter the spatial patterning of post-fire chaparral communities. This study deals with the relationship among fire behavior, post-fire nutrient availabilities and vegetation patterns in a 1-yr old, post-fire Adenostoma fasciculatum chaparral community in the Sierra Nevada Mountains, California, USA. Variations in mineral soil exchangeable cations (Ca, Mg) and extractable phosphorus (P0₄-P) were correlated with ash distribution. Cations and measures of ammonium and nitrate were also correlated with fire intensity, measured by the diameter of the smallest remaining A. fasciculatum twigs following fire. Fire intensity was correlated with the pattern of post-fire vegetation establishment based on first axis DC A scores. However, ash PO4-P was more highly correlated with sample DCA scores, local species richness and total cover (p < 0.01), suggesting that small-scale variations in PO4-P which correlate with ash distributions may be important in structuring this community. Two- and three-term local variance analysis revealed a maximum of pattern intensity in DCA first axis scores at 4–5 m intervals that likely corresponds to pre-fire canopy-gap patterns. However, total cover showed pattern at spatial scales of 8–10 m, and was correlated at this scale with patterns of ash distribution and fire intensity. Microtopographic patterns also occur at similar spatial scales. Microtopographic patterns appear important in determining post-fire plant nutrient and water distributions and, thereby, patterns of plant establishment. Thus, the scale and intensity of post-fire vegetation pattern may differ considerably from pre-fire conditions.     

Prediction of the power of detection of marker-quantitative trait locus linkages using analysis of variance

Analysis of variance can be used to detect the linkage of segregating quantitative trait loci (QTLs) to molecular markers in outbred populations. Using independent full-sib families and assuming linkage equilibrium, equations to predict the power of detection of a QTL are described. These equations are based on an hierarchical analysis of variance assuming either a completely random model or a mixed model, in which the QTL effect is fixed. A simple prediction of power from the mean squares is used that assumes a random model so that in the mixed-model situation this is an approximation. Simulation is used to illustrate the failure of the random model to predict mean squares and, hence, the power. The mixed model is shown to provide accurate prediction of the mean squares and, using the approximation, of power.     

A New Algorithm for Explicit Determination of Variance Component Estimations in Mixed Models and Mixed Classifications of Balanced ANOVA

This paper deals with the balanced case of the analysis of variance. The use of a classification function leads to an easy determination of all possible sources of variation of any mixed classification. For mixed models a new method is derived, which allows to represent explicit the ANOVA-estimations of the variance components respectively the estimation of the mean sum of squares of the fixed effects for all sources of variation. Thereby the corresponding F-quotients and the approximate confidence intervals of variance components are received in a simple way.