Designing occupancy surveys and interpreting non‐detection when observations are imperfect

Aim Conservation practitioners use biological surveys to ascertain whether or not a site is occupied by a particular species. Widely used statistical methods estimate the probability that a species will be detected in a survey of an occupied site. However, these estimates of detection probability are alone not sufficient to calculate the probability that a species is present given that it was not detected. The aim of this paper is to demonstrate methods for correctly calculating (1) the probability a species occupies a site given one or more non‐detections, and (2) the number of sequential non‐detections necessary to assert, with a pre‐specified confidence, that a species is absent from a site. Location Occupancy data for a tree frog in eastern Australia serve to illustrate methods that may be applied anywhere species’ occupancy data are used and detection probabilities are < 1. Methods Building on Bayesian expressions for the probability that a site is occupied by a species when it is not detected, and the number of non‐detections necessary to assert absence with a pre‐specified confidence, we estimate occupancy probabilities across tree frog survey locations, drawing on information about where and when the species was detected during surveys. Results We show that the number of sequential non‐detections necessary to assert that a species is absent increases nonlinearly with the prior probability of occupancy, the probability of detection if present, and the desired level of confidence about absence. Main conclusions If used more widely, the Bayesian analytical approaches illustrated here would improve collection and interpretation of biological survey data, providing a coherent way to incorporate detection probability estimates in the design of minimum survey requirements for monitoring, impact assessment and distribution modelling.     

Bayesian Predictive Probability Functions for Count Data that are Subject to Misclassification

We develop three Bayesian predictive probability functions based on data in the form of a double sample. One Bayesian predictive probability function is for predicting the true unobservable count of interest in a future sample for a Poisson model with data subject to misclassification and two Bayesian predictive probability functions for predicting the number of misclassified counts in a current observable fallible count for an event of interest. We formulate a Gibbs sampler to calculate prediction intervals for these three unobservable random variables and apply our new predictive models to calculate prediction intervals for a real‐data example. (© 2004 WILEY‐VCH Verlag GmbH & Co. KGaA, Weinheim)     

Reanalysis of Murphy et al.'s data gives various mammalian phylogenies and suggests overcredibility of Bayesian trees

Murphy and colleagues reported that the mammalian phylogeny was resolved by Bayesian phylogenetics. However, the DNA sequences they used had many alignment gaps and undetermined nucleotide sites. We therefore reanalyzed their data by minimizing unshared nucleotide sites and retaining as many species as possible (13 species). In constructing phylogenetic trees, we used the Bayesian, maximum likelihood (ML), maximum parsimony (MP), and neighbor-joining (NJ) methods with different substitution models. These trees were constructed by using both protein and DNA sequences. The results showed that the posterior probabilities for Bayesian trees were generally much higher than the bootstrap values for ML, MP, and NJ trees. Two different Bayesian topologies for the same set of species were sometimes supported by high posterior probabilities, implying that two different topologies can be judged to be correct by Bayesian phylogenetics. This suggests that the posterior probability in Bayesian analysis can be excessively high as an indication of statistical confidence and therefore Murphy et al.'s tree, which largely depends on Bayesian posterior probability, may not be correct.     

A Bayesian framework for the analysis of cospeciation

Abstract.— Information on the history of cospeciation and host switching for a group of host and parasite species is contained in the DNA sequences sampled from each. Here, we develop a Bayesian framework for the analysis of cospeciation. We suggest a simple model of host switching by a parasite on a host phylogeny in which host switching events are assumed to occur at a constant rate over the entire evolutionary history of associated hosts and parasites. The posterior probability density of the parameters of the model of host switching are evaluated numerically using Markov chain Monte Carlo. In particular, the method generates the probability density of the number of host switches and of the host switching rate. Moreover, the method provides information on the probability that an event of host switching is associated with a particular pair of branches. A Bayesian approach has several advantages over other methods for the analysis of cospeciation. In particular, it does not assume that the host or parasite phylogenies are known without error; many alternative phylogenies are sampled in proportion to their probability of being correct.     

Associations between anuran tadpoles and salinity in a landscape mosaic of wetlands impacted by secondary salinisation

1. The secondary salinisation of wetlands is a global problem that poses a great threat to most freshwater biodiversity, including amphibians. We examined tadpole diversity in relation to wetland conductivity (our proxy for salinity) in wetlands in south‐eastern Australia to better understand (i) how salinity and amphibian diversity interact and (ii) the threat posed by secondary salinisation. 2. Six tadpole species were trapped in 56 wetlands that reflected a typical salinity gradient for the study region. We developed Bayesian models to examine the relationships between conductivity and both the probability of species occupancy and expected number of species with the imperfect detection probability of species accounted for in the models. 3. The probability of occupancy for all species and expected species number was negatively associated with wetland conductivity. Our results predict that conductivity should not limit tadpole presence below about 3000 μS cm⁻¹ at 25 °C (approximately 6% seawater) in the region, but will largely exclude amphibian larvae beyond about 6000 μS cm⁻¹ at 25 °C (approximately 12% seawater). 4. We also detected subtle among‐species differences in salinity tolerance. The results reported here show that tadpoles in the study region are likely to be negatively affected by projected future increases in salinisation.     

Inferring speciation times under an episodic molecular clock

We extend our recently developed Markov chain Monte Carlo algorithm for Bayesian estimation of species divergence times to allow variable evolutionary rates among lineages. The method can use heterogeneous data from multiple gene loci and accommodate multiple fossil calibrations. Uncertainties in fossil calibrations are described using flexible statistical distributions. The prior for divergence times for nodes lacking fossil calibrations is specified by use of a birth-death process with species sampling. The prior for lineage-specific substitution rates is specified using either a model with autocorrelated rates among adjacent lineages (based on a geometric Brownian motion model of rate drift) or a model with independent rates among lineages specified by a log-normal probability distribution. We develop an infinite-sites theory, which predicts that when the amount of sequence data approaches infinity, the width of the posterior credibility interval and the posterior mean of divergence times form a perfect linear relationship, with the slope indicating uncertainties in time estimates that cannot be reduced by sequence data alone. Simulations are used to study the influence of among-lineage rate variation and the number of loci sampled on the uncertainty of divergence time estimates. The analysis suggests that posterior time estimates typically involve considerable uncertainties even with an infinite amount of sequence data, and that the reliability and precision of fossil calibrations are critically important to divergence time estimation. We apply our new algorithms to two empirical data sets and compare the results with those obtained in previous Bayesian and likelihood analyses. The results demonstrate the utility of our new algorithms.     

Estimating phylogenetic trees from pairwise likelihoods and posterior probabilities of substitution counts

The field of phylogenetic tree estimation has been dominated by three broad classes of methods: distance-based approaches, parsimony and likelihood-based methods (including maximum likelihood (ML) and Bayesian approaches). Here we introduce two new approaches to tree inference: pairwise likelihood estimation and a distance-based method that estimates the number of substitutions along the paths through the tree. Our results include the derivation of the formulae for the probability that two leaves will be identical at a site given a number of substitutions along the path connecting them. We also derive the posterior probability of the number of substitutions along a path between two sequences. The calculations for the posterior probabilities are exact for group-based, symmetric models of character evolution, but are only approximate for more general models.     

On the estimation of species richness based on the accumulation of previously unrecorded species

Estimation of species richness of local communities has become an important topic in community ecology and monitoring. Investigators can seldom enumerate all the species present in the area of interest during sampling sessions. If the location of interest is sampled repeatedly within a short time period, the number of new species recorded is typically largest in the initial sample and decreases as sampling proceeds, but new species may be detected if sampling sessions are added. The question is how to estimate the total number of species. The data collected by sampling the area of interest repeatedly can be used to build species accumulation curves: the cumulative number of species recorded as a function of the number of sampling sessions (which we refer to as “species accumulation data”). A classic approach used to compute total species richness is to fit curves to the data on species accumulation with sampling effort. This approach does not rest on direct estimation of the probability of detecting species during sampling sessions and has no underlying basis regarding the sampling process that gave rise to the data. Here we recommend a probabilistic, nonparametric estimator for species richness for use with species accumulation data. We use estimators of population size that were developed for capture‐recapture data, but that can be used to estimate the size of species assemblages using species accumulation data. Models of detection probability account for the underlying sampling process. They permit variation in detection probability among species. We illustrate this approach using data from the North American Breeding Bird Survey (BBS). We describe other situations where species accumulation data are collected under different designs (e.g., over longer periods of time, or over spatial replicates) and that lend themselves to of use capture‐recapture models for estimating the size of the community of interest. We discuss the assumptions and interpretations corresponding to each situation.     

Decision analysis for conservation breeding: Maximizing production for reintroduction of whooping cranes

Captive breeding is key to management of severely endangered species, but maximizing captive production can be challenging because of poor knowledge of species breeding biology and the complexity of evaluating different management options. In the face of uncertainty and complexity, decision-analytic approaches can be used to identify optimal management options for maximizing captive production. Building decision-analytic models requires iterations of model conception, data analysis, model building and evaluation, identification of remaining uncertainty, further research and monitoring to reduce uncertainty, and integration of new data into the model. We initiated such a process to maximize captive production of the whooping crane (Grus americana), the world's most endangered crane, which is managed through captive breeding and reintroduction. We collected 15 years of captive breeding data from 3 institutions and used Bayesian analysis and model selection to identify predictors of whooping crane hatching success. The strongest predictor, and that with clear management relevance, was incubation environment. The incubation period of whooping crane eggs is split across two environments: crane nests and artificial incubators. Although artificial incubators are useful for allowing breeding pairs to produce multiple clutches, our results indicate that crane incubation is most effective at promoting hatching success. Hatching probability increased the longer an egg spent in a crane nest, from 40% hatching probability for eggs receiving 1 day of crane incubation to 95% for those receiving 30 days (time incubated in each environment varied independently of total incubation period). Because birds will lay fewer eggs when they are incubating longer, a tradeoff exists between the number of clutches produced and egg hatching probability. We developed a decision-analytic model that estimated 16 to be the optimal number of days of crane incubation needed to maximize the number of offspring produced. These results show that using decision-analytic tools to account for uncertainty in captive breeding can improve the rate at which such programs contribute to wildlife reintroductions. © 2011 The Wildlife Society.     

Branch-length prior influences Bayesian posterior probability of phylogeny

The Bayesian method for estimating species phylogenies from molecular sequence data provides an attractive alternative to maximum likelihood with nonparametric bootstrap due to the easy interpretation of posterior probabilities for trees and to availability of efficient computational algorithms. However, for many data sets it produces extremely high posterior probabilities, sometimes for apparently incorrect clades. Here we use both computer simulation and empirical data analysis to examine the effect of the prior model for internal branch lengths. We found that posterior probabilities for trees and clades are sensitive to the prior for internal branch lengths, and priors assuming long internal branches cause high posterior probabilities for trees. In particular, uniform priors with high upper bounds bias Bayesian clade probabilities in favor of extreme values. We discuss possible remedies to the problem, including empirical and full Bayesian methods and subjective procedures suggested in Bayesian hypothesis testing. Our results also suggest that the bootstrap proportion and Bayesian posterior probability are different measures of accuracy, and that the bootstrap proportion, if interpreted as the probability that the clade is true, can be either too liberal or too conservative.     

Bayesian phylogenetic inference using DNA sequences: a Markov Chain Monte Carlo Method

An improved Bayesian method is presented for estimating phylogenetic trees using DNA sequence data. The birth-death process with species sampling is used to specify the prior distribution of phylogenies and ancestral speciation times, and the posterior probabilities of phylogenies are used to estimate the maximum posterior probability (MAP) tree. Monte Carlo integration is used to integrate over the ancestral speciation times for particular trees. A Markov Chain Monte Carlo method is used to generate the set of trees with the highest posterior probabilities. Methods are described for an empirical Bayesian analysis, in which estimates of the speciation and extinction rates are used in calculating the posterior probabilities, and a hierarchical Bayesian analysis, in which these parameters are removed from the model by an additional integration. The Markov Chain Monte Carlo method avoids the requirement of our earlier method for calculating MAP trees to sum over all possible topologies (which limited the number of taxa in an analysis to about five). The methods are applied to analyze DNA sequences for nine species of primates, and the MAP tree, which is identical to a maximum-likelihood estimate of topology, has a probability of approximately 95%.      

The accuracy of species tree estimation under simulation: a comparison of methods

Numerous simulation studies have investigated the accuracy of phylogenetic inference of gene trees under maximum parsimony, maximum likelihood, and Bayesian techniques. The relative accuracy of species tree inference methods under simulation has received less study. The number of analytical techniques available for inferring species trees is increasing rapidly, and in this paper, we compare the performance of several species tree inference techniques at estimating recent species divergences using computer simulation. Simulating gene trees within species trees of different shapes and with varying tree lengths (T) and population sizes (), and evolving sequences on those gene trees, allows us to determine how phylogenetic accuracy changes in relation to different levels of deep coalescence and phylogenetic signal. When the probability of discordance between the gene trees and the species tree is high (i.e., T is small and/or is large), Bayesian species tree inference using the multispecies coalescent (BEST) outperforms other methods. The performance of all methods improves as the total length of the species tree is increased, which reflects the combined benefits of decreasing the probability of discordance between species trees and gene trees and gaining more accurate estimates for gene trees. Decreasing the probability of deep coalescences by reducing also leads to accuracy gains for most methods. Increasing the number of loci from 10 to 100 improves accuracy under difficult demographic scenarios (i.e., coalescent units ≤ 4N(e)), but 10 loci are adequate for estimating the correct species tree in cases where deep coalescence is limited or absent. In general, the correlation between the phylogenetic accuracy and the posterior probability values obtained from BEST is high, although posterior probabilities are overestimated when the prior distribution for is misspecified.     

Tail posterior probability for inference in pairwise and multiclass gene expression data

We consider the problem of identifying differentially expressed genes in microarray data in a Bayesian framework with a noninformative prior distribution on the parameter quantifying differential expression. We introduce a new rule, tail posterior probability, based on the posterior distribution of the standardized difference, to identify genes differentially expressed between two conditions, and we derive a frequentist estimator of the false discovery rate associated with this rule. We compare it to other Bayesian rules in the considered settings. We show how the tail posterior probability can be extended to testing a compound null hypothesis against a class of specific alternatives in multiclass data.     

Bayesian Adaptive Randomization and Trial Monitoring with Predictive Probability for Time-to-Event Endpoint

There has been much development in Bayesian adaptive designs in clinical trials. In the Bayesian paradigm, the posterior predictive distribution characterizes the future possible outcomes given the currently observed data. Based on the interim time-to-event data, we develop a new phase II trial design by combining the strength of both Bayesian adaptive randomization and the predictive probability. By comparing the mean survival times between patients assigned to two treatment arms, more patients are assigned to the better treatment on the basis of adaptive randomization. We continuously monitor the trial using the predictive probability for early termination in the case of superiority or futility. We conduct extensive simulation studies to examine the operating characteristics of four designs: the proposed predictive probability adaptive randomization design, the predictive probability equal randomization design, the posterior probability adaptive randomization design, and the group sequential design. Adaptive randomization designs using predictive probability and posterior probability yield a longer overall median survival time than the group sequential design, but at the cost of a slightly larger sample size. The average sample size using the predictive probability method is generally smaller than that of the posterior probability design.     

Properties of phylogenetic trees generated by Yule-type speciation models

We investigate some discrete structural properties of evolutionary trees generated under simple null models of speciation, such as the Yule model. These models have been used as priors in Bayesian approaches to phylogenetic analysis, and also to test hypotheses concerning the speciation process. In this paper we describe new results for three properties of trees generated under such models. Firstly, for a rooted tree generated by the Yule model we describe the probability distribution on the depth (number of edges from the root) of the most recent common ancestor of a random subset of k species. Next we show that, for trees generated under the Yule model, the approximate position of the root can be estimated from the associated unrooted tree, even for trees with a large number of leaves. Finally, we analyse a biologically motivated extension of the Yule model and describe its distribution on tree shapes when speciation occurs in rapid bursts.     

Bayesian sample size determination using commensurate priors to leverage preexperimental data

This paper develops Bayesian sample size formulae for experiments comparing two groups, where relevant preexperimental information from multiple sources can be incorporated in a robust prior to support both the design and analysis. We use commensurate predictive priors for borrowing of information and further place Gamma mixture priors on the precisions to account for preliminary belief about the pairwise (in)commensurability between parameters that underpin the historical and new experiments. Averaged over the probability space of the new experimental data, appropriate sample sizes are found according to criteria that control certain aspects of the posterior distribution, such as the coverage probability or length of a defined density region. Our Bayesian methodology can be applied to circumstances that compare two normal means, proportions, or event times. When nuisance parameters (such as variance) in the new experiment are unknown, a prior distribution can further be specified based on preexperimental data. Exact solutions are available based on most of the criteria considered for Bayesian sample size determination, while a search procedure is described in cases for which there are no closed-form expressions. We illustrate the application of our sample size formulae in the design of clinical trials, where pretrial information is available to be leveraged. Hypothetical data examples, motivated by a rare-disease trial with an elicited expert prior opinion, and a comprehensive performance evaluation of the proposed methodology are presented.     

Comparing two methods of calculating the probability of absence: the Negative Predictive Value, and a Credible Interval

If sampling fails to reveal the presence of an invasive species with potential to actually be present, how may we calculate the probability that it is truly absent, e.g. didymo (Didymosphenia geminate) in New Zealand’s North Island. In statistical terms this is a Bayesian question, concerning the probability of a hypothesis (presence/absence), given the obtained data (all results negative). “Classical” theory doesn’t answer this question, because it inverts the required considerations by calculating the probability of all samples being absent if the invasive was actually present. Accordingly, the Bayesian view of “probability” must be adopted in order to answer the question. That definition differs from classical probability in that it always includes an element of subjective belief, particularly in the choice of an appropriate “prior probability” (this is our belief as to the presence of the invasive organism before collecting new data). Bayesian methods can therefore be somewhat controversial – but we seem forced to use them. One Bayesian approach is to use the “Negative Predictive Value”, in which a point estimate of the probability of presence (or absence) prior to sample collection (the “prior probability”) is updated using data once collected using Bayes’ rule. This is in common use in medical studies, where the prior probability is the background disease prevalence, which is generally well understood. It is sometimes used in environmental ‘hot-spot’ investigations. An alternative approach is to recognise the uncertainty in the prior belief by using a distribution of prior probability and updating that using data once collected to give a Credible Interval in which the probability of presence (or absence) should lie – if all our assumptions have been satisfied. We will compare the merits of these approaches considering didymo, southern salt marsh mosquito (Aedes camptorhychus) and the sea squirt Styela clava.     

Hierarchical Bayesian integrated model for estimating migratory bird harvest in Canada

The Canadian Wildlife Service (CWS) requires reliable estimates of the harvest of migratory game birds, including waterfowl, to effectively manage populations of these hunted species. The National Harvest Survey is an annual survey of hunters who purchase Canada's mandatory migratory game bird hunting permit, integrating information from a survey of hunting activity with information from a separate survey of species composition in the harvest. We used these survey data to estimate the number of birds harvested for each species and hunting activity metrics (e.g., number of active hunters, days spent hunting). The analytical methods used to generate these estimates have not changed since the survey was first designed in the early 1970s. We describe a new hierarchical Bayesian integrated model, which replaces the series of ratio estimators that comprised the old model. We are using this new model to generate estimates for migratory bird harvests as of the 2019–2020 hunting season, and to generate updated estimates for all earlier years. The hierarchical Bayesian model uses over-dispersed Poisson distributions to model mean hunter activity and harvest (zero-inflated Poisson and zero-truncated Poisson, respectively). It also includes multinomial distributions to model some key components (e.g., variation in harvest across periods of the hunting season, the species composition of the harvest within each of those periods, the age and sex composition in the harvests of a given species). We estimated the parameters of the Poisson and the multinomial distributions for each year as random effects using first-difference time-series. This time-series component allows the model to share information across years and reduces the sensitivity of the estimates to annual sampling noise. The new model estimates are generally very similar to those from the old model, particularly for the species that occur most commonly in the harvest, so the results do not suggest any major changes to harvest management decisions and regulations. Estimates for all species from the new model are more precise and less susceptible to annual sampling error, particularly for species that occur less commonly in the harvest (e.g., sea ducks, other species of conservation concern). This new model, with its hierarchical Bayesian framework, will also facilitate future improvements and elaborations, allowing the incorporation of prior information from the rich literature and knowledge in game bird management and biology.     

Frequentist properties of Bayesian posterior probabilities of phylogenetic trees under simple and complex substitution models

What does the posterior probability of a phylogenetic tree mean?This simulation study shows that Bayesian posterior probabilities have the meaning that is typically ascribed to them; the posterior probability of a tree is the probability that the tree is correct, assuming that the model is correct. At the same time, the Bayesian method can be sensitive to model misspecification, and the sensitivity of the Bayesian method appears to be greater than the sensitivity of the nonparametric bootstrap method (using maximum likelihood to estimate trees). Although the estimates of phylogeny obtained by use of the method of maximum likelihood or the Bayesian method are likely to be similar, the assessment of the uncertainty of inferred trees via either bootstrapping (for maximum likelihood estimates) or posterior probabilities (for Bayesian estimates) is not likely to be the same. We suggest that the Bayesian method be implemented with the most complex models of those currently available, as this should reduce the chance that the method will concentrate too much probability on too few trees.     

A model-based method for identifying species hybrids using multilocus genetic data

We present a statistical method for identifying species hybrids using data on multiple, unlinked markers. The method does not require that allele frequencies be known in the parental species nor that separate, pure samples of the parental species be available. The method is suitable for both markers with fixed allelic differences between the species and markers without fixed differences. The probability model used is one in which parentals and various classes of hybrids (F(1)'s, F(2)'s, and various backcrosses) form a mixture from which the sample is drawn. Using the framework of Bayesian model-based clustering allows us to compute, by Markov chain Monte Carlo, the posterior probability that each individual belongs to each of the distinct hybrid classes. We demonstrate the method on allozyme data from two species of hybridizing trout, as well as on two simulated data sets.