Phylogenetically nested comparisons for testing correlates of species richness: a simulation study of continuous variables

Abstract.— Explaining the uneven distribution of species among lineages is one of the oldest questions in evolution. Proposed correlations between biological traits and species diversity are routinely tested by making comparisons between phylogenetic sister clades. Several recent studies have used nested sister-clade comparisons to test hypotheses linking continuously varying traits, such as body size, with diversity. Evaluating the findings of these studies is complicated because they differ in the index of species richness difference used, the way in which trait differences were treated, and the statistical tests employed. In this paper, we use simulations to compare the performance of four species richness indices, two choices about the branch lengths used to estimate trait values for internal nodes and two statistical tests under a range of models of clade growth and character evolution. All four indices returned appropriate Type I error rates when the assumptions of the method were met and when branch lengths were set proportional to time. Only two of the indices were robust to the different evolutionary models and to different choices of branch lengths and statistical tests. These robust indices had comparable power under one nonnull scenario. Regression through the origin was consistently more powerful than the t -test, and the choice of branch lengths exerts a strong effect on both the validity and power. In the light of our simulations, we re-evaluate the findings of those who have previously used nested comparisons in the context of species richness. We provide a set of simple guidelines to maximize the performance of phylogenetically nested comparisons in tests of putative correlates of species richness.     

Effects of branch length errors on the performance of phylogenetically independent contrasts

We examined Type I error rates of Felsenstein's (1985; Am. Nat. 125:1-15) comparative method of phylogenetically independent contrasts when branch lengths are in error and the model of evolution is not Brownian motion. We used seven evolutionary models, six of which depart strongly from Brownian motion, to simulate the evolution of two continuously valued characters along two different phylogenies (15 and 49 species). First, we examined the performance of independent contrasts when branch lengths are distorted systematically, for example, by taking the square root of each branch segment. These distortions often caused inflated Type I error rates, but performance was almost always restored when branch length transformations were used. Next, we investigated effects of random errors in branch lengths. After the data were simulated, we added errors to the branch lengths and then used the altered phylogenies to estimate character correlations. Errors in the branches could be of two types: fixed, where branch lengths are either shortened or lengthened by a fixed fraction; or variable, where the error is a normal variate with mean zero and the variance is scaled to the length of the branch (so that expected error relative to branch length is constant for the whole tree). Thus, the error added is unrelated to the microevolutionary model. Without branch length checks and transformations, independent contrasts tended to yield extremely inflated and highly variable Type I error rates. Type I error rates were reduced, however, when branch lengths were checked and transformed as proposed by Garland et al. (1992; Syst. Biol. 41:18-32), and almost never exceeded twice the nominal P-value at alpha = 0.05. Our results also indicate that, if branch length transformations are applied, then the appropriate degrees of freedom for testing the significance of a correlation coefficient should, in general, be reduced to account for estimation of the best branch length transformation. These results extend those reported in Díaz-Uriarte and Garland (1996; Syst. Biol. 45:27-47), and show that, even with errors in branch lengths and evolutionary models different from Brownian motion, independent contrasts are a robust method for testing hypotheses of correlated evolution.     

Investigating evolutionary lag using the species-pairs evolutionary lag test (SPELT)

For traits showing correlated evolution, one trait may evolve more slowly than the other, producing evolutionary lag. The species-pairs evolutionary lag test (SPELT) uses an independent contrasts based approach to detect evolutionary lag on a phylogeny. We investigated the statistical performance of SPELT in relation to degree of lag, sample size (species pairs), and strength of association between traits. We simulated trait evolution under two models: one in which trait X changes during speciation and the lagging trait Y catches up as a function of time since speciation; and another in which trait X evolves in a random walk and the lagging trait Y is a function of X at a previous time period. Type I error rates under “no lag” were close to the expected level of 5%, indicating that the method is not prone to false-positives. Simulation results suggest that reasonable statistical power (80%) is reached with around 140 species pairs, although the degree of lag and trait associations had additional influences on power. We applied the method to two datasets and discuss how estimation of a branch length scaling parameter (κ) can be used with SPELT to detect lag.     

PHYLOGENETIC ANALYSES OF THE CORRELATED EVOLUTION OF CONTINUOUS CHARACTERS: A SIMULATION STUDY

We use computer simulation to compare the statistical properties of several methods that have been proposed for estimating the evolutionary correlation between two continuous traits, and define alternative evolutionary correlations that may be of interest. We focus on Felsenstein's (1985) method and some variations of it and on several “minimum evolution” methods (of which the procedure of Huey and Bennett [1987] is a special case), as compared with a nonphylogenetic correlation. The last, a simple correlation of trait values across the tips of a phylogeny, virtually always yields inflated Type I error rates, relatively low power, and relatively poor estimates of evolutionary correlations. We therefore cannot recommend its use. In contrast, Felsenstein's (1985) method yields acceptable significance tests, high power, and good estimates of what we term the input correlation and the standardized realized evolutionary correlation, given complete phylogenetic information and knowledge of the rate and mode of character change (e.g., gradual and proportional to time [“Brownian motion”] or punctuational, with change only at speciation events). Inaccurate branch length information may affect any method adversely, but only rarely does it cause Felsenstein's (1985) method to perform worse than do the others tested. Other proposed methods generally yield inflated Type I error rates and have lower power. However, certain minimum evolution methods (although not the specific procedure used by Huey and Bennett [1987]) often provide more accurate estimates of what we term the unstandardized realized evolutionary correlation, and their use is recommended when estimation of this correlation is desired. We also demonstrate how correct Type I error rates can be obtained for any method by reference to an empirical null distribution derived from computer simulations, and provide practical suggestions on choosing an analytical method, based both on the evolutionary correlation of interest and on the availability of branch lengths and knowledge of the model of evolutionary change appropriate for the characters being analyzed. Computer programs that implement the various methods and that will simulate (correlated) character evolution along a known phylogeny are available from the authors on request. These programs can be used to test the effectiveness of any new methods that might be proposed, and to check the generality of our conclusions with regard to other phylogenies.     

A METHOD FOR ASSESSING PHYLOGENETIC LEAST SQUARES MODELS FOR SHAPE AND OTHER HIGH‐DIMENSIONAL MULTIVARIATE DATA

Studies of evolutionary correlations commonly use phylogenetic regression (i.e., independent contrasts and phylogenetic generalized least squares) to assess trait covariation in a phylogenetic context. However, while this approach is appropriate for evaluating trends in one or a few traits, it is incapable of assessing patterns in highly multivariate data, as the large number of variables relative to sample size prohibits parametric test statistics from being computed. This poses serious limitations for comparative biologists, who must either simplify how they quantify phenotypic traits, or alter the biological hypotheses they wish to examine. In this article, I propose a new statistical procedure for performing ANOVA and regression models in a phylogenetic context that can accommodate high‐dimensional datasets. The approach is derived from the statistical equivalency between parametric methods using covariance matrices and methods based on distance matrices. Using simulations under Brownian motion, I show that the method displays appropriate Type I error rates and statistical power, whereas standard parametric procedures have decreasing power as data dimensionality increases. As such, the new procedure provides a useful means of assessing trait covariation across a set of taxa related by a phylogeny, enabling macroevolutionary biologists to test hypotheses of adaptation, and phenotypic change in high‐dimensional datasets.     

Correlated evolution in fig pollination

This is the first comparative study of correlated evolution between figs (Ficus species, Moraceae) and their pollinators (Hymenoptera: Agaoninae) based on molecular phylogenies of both lineages. Fig relationships based on the internal transcribed spacer region (ITS) of nuclear ribosomal DNA and pollinator relationships inferred from mitochondrial cytochrome oxidase I (COI) sequences enabled the study of correlated evolution based on molecular phylogenies for the largest set of interacting species ever compared. Comparative methods have been applied to tests of adaptation, but the application of these methods in tests of coadaptation, defined as reciprocal evolutionary change in interacting lineages, has received less attention. I have extended tests of correlated evolution between two traits along a phylogeny to the case of interacting lineages, where two traits may or may not share a common phylogenetic history. Independent contrasts and phylogenetic autocorrelation rejected the null hypothesis that trait correlations within lineages are stronger than trait correlations between interacting lineages. Fig style lengths and pollinator ovipositor lengths, for example, were more highly correlated than were pollinator body size and ovipositor length. Mutualistic interactions between figs and their pollinators illustrate the novel ways in which phylogenies and comparative methods can detect patterns of correlated evolution. The most outstanding evidence of correlated evolution between these obligate mutualists is that interacting trait correlations are stronger than within-lineage allometric relationships.     

TESTING FOR PHYLOGENETIC SIGNAL IN BIOLOGICAL TRAITS: THE UBIQUITY OF CROSS‐PRODUCT STATISTICS

To evaluate rates of evolution, to establish tests of correlation between two traits, or to investigate to what degree the phylogeny of a species assemblage is predictive of a trait value so‐called tests for phylogenetic signal are used. Being based on different approaches, these tests are generally thought to possess quite different statistical performances. In this article, we show that the Blomberg et al. K and K*, the Abouheif index, the Moran's I, and the Mantel correlation are all based on a cross‐product statistic, and are thus all related to each other when they are associated to a permutation test of phylogenetic signal. What changes is only the way phylogenetic and trait similarities (or dissimilarities) among the tips of a phylogeny are computed. The definitions of the phylogenetic and trait‐based (dis)similarities among tips thus determines the performance of the tests. We shortly discuss the biological and statistical consequences (in terms of power and type I error of the tests) of the observed relatedness among the statistics that allow tests for phylogenetic signal. Blomberg et al. K* statistic appears as one on the most efficient approaches to test for phylogenetic signal. When branch lengths are not available or not accurate, Abouheif's C_mean statistic is a powerful alternative to K*.     

Body mass and foraging ecology predict evolutionary patterns of skeletal pneumaticity in the diverse "waterbird" clade

Extensive skeletal pneumaticity (air-filled bone) is a distinguishing feature of birds. The proportion of the skeleton that is pneumatized varies considerably among the >10,000 living species, with notable patterns including increases in larger bodied forms, and reductions in birds employing underwater pursuit diving as a foraging strategy. I assess the relationship between skeletal pneumaticity and body mass and foraging ecology, using a dataset of the diverse "waterbird" clade that encompasses a broad range of trait variation. Inferred changes in pneumaticity and body mass are congruent across different estimates of phylogeny, whereas pursuit diving has evolved independently between two and five times. Phylogenetic regressions detected positive relationships between body mass and pneumaticity, and negative relationships between pursuit diving and pneumaticity, whether independent variables are considered in isolation or jointly. Results are generally consistent across different estimates of topology and branch lengths. "Predictive" analyses reveal that several pursuit divers (loons, penguins, cormorants, darters) are significantly apneumatic compared to their relatives, and provide an example of how phylogenetic information can increase the statistical power to detect taxa that depart from established trait correlations. These findings provide the strongest quantitative comparative support yet for classical hypotheses regarding the evolution of avian skeletal pneumaticity.     

Assessing Trait Covariation and Morphological Integration on Phylogenies Using Evolutionary Covariance Matrices

Morphological integration describes the degree to which sets of organismal traits covary with one another. Morphological covariation may be evaluated at various levels of biological organization, but when characterizing such patterns across species at the macroevolutionary level, phylogeny must be taken into account. We outline an analytical procedure based on the evolutionary covariance matrix that allows species-level patterns of morphological integration among structures defined by sets of traits to be evaluated while accounting for the phylogenetic relationships among taxa, providing a flexible and robust complement to related phylogenetic independent contrasts based approaches. Using computer simulations under a Brownian motion model we show that statistical tests based on the approach display appropriate Type I error rates and high statistical power for detecting known levels of integration, and these trends remain consistent for simulations using different numbers of species, and for simulations that differ in the number of trait dimensions. Thus, our procedure provides a useful means of testing hypotheses of morphological integration in a phylogenetic context. We illustrate the utility of this approach by evaluating evolutionary patterns of morphological integration in head shape for a lineage of Plethodon salamanders, and find significant integration between cranial shape and mandible shape. Finally, computer code written in R for implementing the procedure is provided.     

Effects of taxon sampling and tree reconstruction methods on phylodiversity metrics

相似文献   

A likelihood-based method for testing for nonstochastic variation of diversification rates in phylogenies

Observed variations in rates of taxonomic diversification have been attributed to a range of factors including biological innovations, ecosystem restructuring, and environmental changes. Before inferring causality of any particular factor, however, it is critical to demonstrate that the observed variation in diversity is significantly greater than that expected from natural stochastic processes. Relative tests that assess whether observed asymmetry in species richness between sister taxa in monophyletic pairs is greater than would be expected under a symmetric model have been used widely in studies of rate heterogeneity and are particularly useful for groups in which paleontological data are problematic. Although one such test introduced by Slowinski and Guyer a decade ago has been applied to a wide range of clades and evolutionary questions, the statistical behavior of the test has not been examined extensively, particularly when used with Fisher's procedure for combining probabilities to analyze data from multiple independent taxon pairs. Here, certain pragmatic difficulties with the Slowinski-Guyer test are described, further details of the development of a recently introduced likelihood-based relative rates test are presented, and standard simulation procedures are used to assess the behavior of the two tests in a range of situations to determine: (1) the accuracy of the tests' nominal Type I error rate; (2) the statistical power of the tests; (3) the sensitivity of the tests to inclusion of taxon pairs with few species; (4) the behavior of the tests with datasets comprised of few taxon pairs; and (5) the sensitivity of the tests to certain violations of the null model assumptions. Our results indicate that in most biologically plausible scenarios, the likelihood-based test has superior statistical properties in terms of both Type I error rate and power, and we found no scenario in which the Slowinski-Guyer test was distinctly superior, although the degree of the discrepancy varies among the different scenarios. The Slowinski-Guyer test tends to be much more conservative (i.e., very disinclined to reject the null hypothesis) in datasets with many small pairs. In most situations, the performance of both the likelihood-based test and particularly the Slowinski-Guyer test improve when pairs with few species are excluded from the computation, although this is balanced against a decline in the tests' power and accuracy as fewer pairs are included in the dataset. The performance of both tests is quite poor when they are applied to datasets in which the taxon sizes do not conform to the distribution implied by the usual null model. Thus, results of analyses of taxonomic rate heterogeneity using the Slowinski-Guyer test can be misleading because the test's ability to reject the null hypothesis (equal rates) when true is often inaccurate and its ability to reject the null hypothesis when the alternative (unequal rates) is true is poor, particularly when small taxon pairs are included. Although not always perfect, the likelihood-based test provides a more accurate and powerful alternative as a relative rates test.     

FiSSE: A simple nonparametric test for the effects of a binary character on lineage diversification rates

It is widely assumed that phenotypic traits can influence rates of speciation and extinction, and several statistical approaches have been used to test for correlations between character states and lineage diversification. Recent work suggests that model‐based tests of state‐dependent speciation and extinction are sensitive to model inadequacy and phylogenetic pseudoreplication. We describe a simple nonparametric statistical test (“FiSSE”) to assess the effects of a binary character on lineage diversification rates. The method involves computing a test statistic that compares the distributions of branch lengths for lineages with and without a character state of interest. The value of the test statistic is compared to a null distribution generated by simulating character histories on the observed phylogeny. Our tests show that FiSSE can reliably infer trait‐dependent speciation on phylogenies of several hundred tips. The method has low power to detect trait‐dependent extinction but can infer state‐dependent differences in speciation even when net diversification rates are constant. We assemble a range of macroevolutionary scenarios that are problematic for likelihood‐based methods, and we find that FiSSE does not show similarly elevated false positive rates. We suggest that nonparametric statistical approaches, such as FiSSE, provide an important complement to formal process‐based models for trait‐dependent diversification.     

Positive correlations between molecular and morphological rates of evolution

The existence of positive associations between rates of molecular and morphological evolution (calculated from branch lengths of phylogenetic trees reconstructed using molecular and morphological characters, respectively) is important to issues of neutrality in sequence evolution, phylogenetic reconstructions assuming neutrality, and evolutionary genotype-phenotype mapping. Rates correlate positively when including branches leading to extant species (tips). Excluding tips, trends are similar, but statistical significances decrease systematically. This is due to (a) lower statistical power (excluding tips reduces sample sizes), and (b) rates are solely calculated from inaccurately reconstructed character states of extinct ancestral species, and this noise decreases correlation strengths. Correlations between molecular and morphological rates of evolution increase as more morphological characters are included for phylogenetic reconstruction. Sequence lengths apparently affect correlations along similar principles. Analyses of plant phylogenies confirm those from animals: sampling biases decrease correlations between molecular and morphological rates of evolution. Results confirm that genotype and phenotype are linked, and suggest adaptive components for molecular evolution. The discussion stresses the difficulties associated with analyses and conclusions based on data deduced from phylogenetic reconstruction.     

Permutation tests for phylogenetic comparative analyses of high‐dimensional shape data: What you shuffle matters

Evaluating statistical trends in high‐dimensional phenotypes poses challenges for comparative biologists, because the high‐dimensionality of the trait data relative to the number of species can prohibit parametric tests from being computed. Recently, two comparative methods were proposed to circumvent this difficulty. One obtains phylogenetic independent contrasts for all variables, and statistically evaluates the linear model by permuting the phylogenetically independent contrasts (PICs) of the response data. The other uses a distance‐based approach to obtain coefficients for generalized least squares models (D‐PGLS), and subsequently permutes the original data to evaluate the model effects. Here, we show that permuting PICs is not equivalent to permuting the data prior to the analyses as in D‐PGLS. We further explain why PICs are not the correct exchangeable units under the null hypothesis, and demonstrate that this misspecification of permutable units leads to inflated type I error rates of statistical tests. We then show that simply shuffling the original data and recalculating the independent contrasts with each iteration yields significance levels that correspond to those found using D‐PGLS. Thus, while summary statistics from methods based on PICs and PGLS are the same, permuting PICs can lead to strikingly different inferential outcomes with respect to statistical and biological inferences.     

Phylogenetic extinction rates and comparative methodology

Species are not independent points for comparative analyses because closely related species share more evolutionary history and are therefore more similar to each other than distantly related species. The extent to which independent-contrast analysis reduces type I and type II statistical error in comparison with cross-species analysis depends on the relative branch lengths in the phylogenetic tree: as deeper branches get relatively long, cross-species analyses have more statistical type I and type II error. Phylogenetic trees reconstructed from extant species, under the assumptions of a branching process with speciation (branching) and extinction rates remaining constant through time, will have relatively longer deep branches as the extinction rate increases relative to the speciation rate. We compare the statistical performance of cross-species and independent-contrast analyses with varying relative extinction rates, and conclude that cross-species comparisons have unacceptable statistical performance, particularly when extinction rates are relatively high.     

Assessment of the relative merits of a few methods to detect evolutionary trends

Some of the most basic questions about the history of life concern evolutionary trends. These include determining whether or not metazoans have become more complex over time, whether or not body size tends to increase over time (the Cope-Depéret rule), or whether or not brain size has increased over time in various taxa, such as mammals and birds. Despite the proliferation of studies on such topics, assessment of the reliability of results in this field is hampered by the variability of techniques used and the lack of statistical validation of these methods. To solve this problem, simulations are performed using a variety of evolutionary models (gradual Brownian motion, speciational Brownian motion, and Ornstein-Uhlenbeck), with or without a drift of variable amplitude, with variable variance of tips, and with bounds placed close or far from the starting values and final means of simulated characters. These are used to assess the relative merits (power, Type I error rate, bias, and mean absolute value of error on slope estimate) of several statistical methods that have recently been used to assess the presence of evolutionary trends in comparative data. Results show widely divergent performance of the methods. The simple, nonphylogenetic regression (SR) and variance partitioning using phylogenetic eigenvector regression (PVR) with a broken stick selection procedure have greatly inflated Type I error rate (0.123-0.180 at a 0.05 threshold), which invalidates their use in this context. However, they have the greatest power. Most variants of Felsenstein's independent contrasts (FIC; five of which are presented) have adequate Type I error rate, although two have a slightly inflated Type I error rate with at least one of the two reference trees (0.064-0.090 error rate at a 0.05 threshold). The power of all contrast-based methods is always much lower than that of SR and PVR, except under Brownian motion with a strong trend and distant bounds. Mean absolute value of error on slope of all FIC methods is slightly higher than that of phylogenetic generalized least squares (PGLS), SR, and PVR. PGLS performs well, with low Type I error rate, low error on regression coefficient, and power comparable with some FIC methods. Four variants of skewness analysis are examined, and a new method to assess significance of results is presented. However, all have consistently low power, except in rare combinations of trees, trend strength, and distance between final means and bounds. Globally, the results clearly show that FIC-based methods and PGLS are globally better than nonphylogenetic methods and variance partitioning with PVR. FIC methods and PGLS are sensitive to the model of evolution (and, hence, to branch length errors). Our results suggest that regressing raw character contrasts against raw geological age contrasts yields a good combination of power and Type I error rate. New software to facilitate batch analysis is presented.     

POOR STATISTICAL PERFORMANCE OF THE MANTEL TEST IN PHYLOGENETIC COMPARATIVE ANALYSES

The Mantel test, based on comparisons of distance matrices, is commonly employed in comparative biology, but its statistical properties in this context are unknown. Here, we evaluate the performance of the Mantel test for two applications in comparative biology: testing for phylogenetic signal, and testing for an evolutionary correlation between two characters. We find that the Mantel test has poor performance compared to alternative methods, including low power and, under some circumstances, inflated type‐I error. We identify a remedy for the inflated type‐I error of three‐way Mantel tests using phylogenetic permutations; however, this test still has considerably lower power than independent contrasts. We recommend that use of the Mantel test should be restricted to cases in which data can only be expressed as pairwise distances among taxa.     

On the analysis of phylogenetically paired designs

As phylogenetically controlled experimental designs become increasingly common in ecology, the need arises for a standardized statistical treatment of these datasets. Phylogenetically paired designs circumvent the need for resolved phylogenies and have been used to compare species groups, particularly in the areas of invasion biology and adaptation. Despite the widespread use of this approach, the statistical analysis of paired designs has not been critically evaluated. We propose a mixed model approach that includes random effects for pair and species. These random effects introduce a “two-layer” compound symmetry variance structure that captures both the correlations between observations on related species within a pair as well as the correlations between the repeated measurements within species. We conducted a simulation study to assess the effect of model misspecification on Type I and II error rates. We also provide an illustrative example with data containing taxonomically similar species and several outcome variables of interest. We found that a mixed model with species and pair as random effects performed better in these phylogenetically explicit simulations than two commonly used reference models (no or single random effect) by optimizing Type I error rates and power. The proposed mixed model produces acceptable Type I and II error rates despite the absence of a phylogenetic tree. This design can be generalized to a variety of datasets to analyze repeated measurements in clusters of related subjects/species.     

Factors influencing changes in trait correlations across species after using phylogenetic independent contrasts

Comparative interspecific data sets have been analyzed routinely by phylogenetic methods, generally using Felsenstein’s phylogenetic independent contrasts (PIC) method. However, some authors have suggested that it may not be always necessary to incorporate phylogenetic information into statistical analyses of comparative data due to the low influence of shared history on the distribution of␣character states. The main goal of this paper was to undertake a comparison of results from non-phylogenetic Pearson correlation of tip values (TIPs) and phylogenetic independent contrasts analyses (PICs), using 566 correlation coefficients derived from 65 published papers. From each study we collected the following data: taxonomic group, number of species, type of phylogeny, number of polytomies in the phylogenetic tree, if branch length was transformed or not, trait types, the original correlation coefficient between the traits (TIPs) and the correlation coefficient between the traits using the independent contrasts method (PICs). The slope estimated from a regression of PIC correlations on TIP correlations was lower than one, and a paired t-test showed that correlations from PIC are significantly smaller than those obtained by TIP. Thus, PIC analyses tend to decrease the correlation between traits and usually increases the P-value and, thus, favoring the acceptance of the null hypothesis. Multiple factors, including taxonomic group, trait type and use of branch length transformations affected the change in decision regarding the acceptance of the null hypotheses and differences between PIC and TIP results. Due to the variety of factors affecting the differences between results provided by these methods, we suggest that comparative methods should be applied as a conservative approach to cross-species studies. Despite difficulties in quantifying precisely why these factors affect the differences between PIC and TIP, we also suggest that a better evaluation of evolutionary models underlying trait evolution is still necessary in this context and might explain some of the observed patterns.