Five (or so) challenges for species distribution modelling

Species distribution modelling is central to both fundamental and applied research in biogeography. Despite widespread use of models, there are still important conceptual ambiguities as well as biotic and algorithmic uncertainties that need to be investigated in order to increase confidence in model results. We identify and discuss five areas of enquiry that are of high importance for species distribution modelling: (1) clarification of the niche concept; (2) improved designs for sampling data for building models; (3) improved parameterization; (4) improved model selection and predictor contribution; and (5) improved model evaluation. The challenges discussed in this essay do not preclude the need for developments of other areas of research in this field. However, they are critical for allowing the science of species distribution modelling to move forward.

Measurement Errors Should Always Be Incorporated in Phylogenetic Comparative Analysis

The evolution of continuous traits is the central component of comparative analyses in phylogenetics, and the comparison of alternative models of trait evolution has greatly improved our understanding of the mechanisms driving phenotypic differentiation. Several factors influence the comparison of models, and we explore the effects of random errors in trait measurement on the accuracy of model selection. We simulate trait data under a Brownian motion model (BM) and introduce different magnitudes of random measurement error. We then evaluate the resulting statistical support for this model against two alternative models: Ornstein-Uhlenbeck (OU) and accelerating/decelerating rates (ACDC). Our analyses show that even small measurement errors (10%) consistently bias model selection towards erroneous rejection of BM in favour of more parameter-rich models (most frequently the OU model). Fortunately, methods that explicitly incorporate measurement errors in phylogenetic analyses considerably improve the accuracy of model selection. Our results call for caution in interpreting the results of model selection in comparative analyses, especially when complex models garner only modest additional support. Importantly, as measurement errors occur in most trait data sets, we suggest that estimation of measurement errors should always be performed during comparative analysis to reduce chances of misidentification of evolutionary processes.

Regression analysis of spatial data.

Many of the most interesting questions ecologists ask lead to analyses of spatial data. Yet, perhaps confused by the large number of statistical models and fitting methods available, many ecologists seem to believe this is best left to specialists. Here, we describe the issues that need consideration when analysing spatial data and illustrate these using simulation studies. Our comparative analysis involves using methods including generalized least squares, spatial filters, wavelet revised models, conditional autoregressive models and generalized additive mixed models to estimate regression coefficients from synthetic but realistic data sets, including some which violate standard regression assumptions. We assess the performance of each method using two measures and using statistical error rates for model selection. Methods that performed well included generalized least squares family of models and a Bayesian implementation of the conditional auto-regressive model. Ordinary least squares also performed adequately in the absence of model selection, but had poorly controlled Type I error rates and so did not show the improvements in performance under model selection when using the above methods. Removing large-scale spatial trends in the response led to poor performance. These are empirical results; hence extrapolation of these findings to other situations should be performed cautiously. Nevertheless, our simulation-based approach provides much stronger evidence for comparative analysis than assessments based on single or small numbers of data sets, and should be considered a necessary foundation for statements of this type in future.

To select or to weigh: a comparative study of linear combination schemes for superparent-one-dependence estimators

We conduct a large-scale comparative study on linearly combining superparent-one-dependence estimators (SPODEs), a popular family of seminaive Bayesian classifiers. Altogether, 16 model selection and weighing schemes, 58 benchmark data sets, and various statistical tests are employed. This paper's main contributions are threefold. First, it formally presents each scheme's definition, rationale, and time complexity and hence can serve as a comprehensive reference for researchers interested in ensemble learning. Second, it offers bias-variance analysis for each scheme's classification error performance. Third, it identifies effective schemes that meet various needs in practice. This leads to accurate and fast classification algorithms which have an immediate and significant impact on real-world applications. Another important feature of our study is using a variety of statistical tests to evaluate multiple learning methods across multiple data sets.

Multi-source composite kernels for urban image classification

This letter presents advanced classification methods for very high resolution images. Efficient multisource information, both spectral and spatial, is exploited through the use of composite kernels in support vector machines. Weighted summations of kernels accounting for separate sources of spectral and spatial information are analyzed and compared to classical approaches such as pure spectral classification or stacked approaches using all the features in a single vector. Model selection problems are addressed, as well as the importance of the different kernels in the weighted summation.

Effect of cattle on Salmonella carriage, diversity and antimicrobial resistance in free-ranging wild boar (Sus scrofa) in northeastern Spain

Salmonella is distributed worldwide and is a pathogen of economic and public health importance. As a multi-host pathogen with a long environmental persistence, it is a suitable model for the study of wildlife-livestock interactions. In this work, we aim to explore the spill-over of Salmonella between free-ranging wild boar and livestock in a protected natural area in NE Spain and the presence of antimicrobial resistance. Salmonella prevalence, serotypes and diversity were compared between wild boars, sympatric cattle and wild boars from cattle-free areas. The effect of age, sex, cattle presence and cattle herd size on Salmonella probability of infection in wild boars was explored by means of Generalized Linear Models and a model selection based on the Akaike’s Information Criterion. Prevalence was higher in wild boars co-habiting with cattle (35.67%, CI 95% 28.19–43.70) than in wild boar from cattle-free areas (17.54%, CI 95% 8.74–29.91). Probability of a wild boar being a Salmonella carrier increased with cattle herd size but decreased with the host age. Serotypes Meleagridis, Anatum and Othmarschen were isolated concurrently from cattle and sympatric wild boars. Apart from serotypes shared with cattle, wild boars appear to have their own serotypes, which are also found in wild boars from cattle-free areas (Enteritidis, Mikawasima, 4:b:- and 35:r:z35). Serotype richness (diversity) was higher in wild boars co-habiting with cattle, but evenness was not altered by the introduction of serotypes from cattle. The finding of a S. Mbandaka strain resistant to sulfamethoxazole, streptomycin and chloramphenicol and a S. Enteritidis strain resistant to ciprofloxacin and nalidixic acid in wild boars is cause for public health concern.

Falsification and corroboration of conceptual hydrological models using geophysical data

Geophysical data may provide crucial information about hydrological properties, states, and processes that are difficult to obtain by other means. Large data sets can be acquired over widely different scales in a minimally invasive manner and at comparatively low costs, but their effective use in hydrology makes it necessary to understand the fidelity of geophysical models, the assumptions made in their construction, and the links between geophysical and hydrological properties. Geophysics has been applied for groundwater prospecting for almost a century, but it is only in the last 20 years that it is regularly used together with classical hydrological data to build predictive hydrological models. A largely unexplored venue for future work is to use geophysical data to falsify or rank competing conceptual hydrological models. A promising cornerstone for such a model selection strategy is the Bayes factor, but it can only be calculated reliably when considering the main sources of uncertainty throughout the hydrogeophysical parameter estimation process. Most classical geophysical imaging tools tend to favor models with smoothly varying property fields that are at odds with most conceptual hydrological models of interest. It is thus necessary to account for this bias or use alternative approaches in which proposed conceptual models are honored at all steps in the model building process.

Adaptive MCMC methods with applications in environmental and geophysical models

This work presents new, efficient Markov chain Monte Carlo (MCMC) simulation methods for statistical analysis in various modelling applications. When using MCMC methods, the model is simulated repeatedly to explore the probability distribution describing the uncertainties in model parameters and predictions. In adaptive MCMC methods based on the Metropolis-Hastings algorithm, the proposal distribution needed by the algorithm learns from the target distribution as the simulation proceeds. Adaptive MCMC methods have been subject of intensive research lately, as they open a way for essentially easier use of the methodology. The lack of user-friendly computer programs has been a main obstacle for wider acceptance of the methods. This work provides two new adaptive MCMC methods: DRAM and AARJ. The DRAM method has been built especially to work in high dimensional and non-linear problems. The AARJ method is an extension to DRAM for model selection problems, where the mathematical formulation of the model is uncertain and we want simultaneously to fit several different models to the same observations. The methods were developed while keeping in mind the needs of modelling applications typical in environmental sciences. The development work has been pursued while working with several application projects. The applications presented in this work are: a winter time oxygen concentration model for Lake Tuusulanjärvi and adaptive control of the aerator; a nutrition model for Lake Pyhäjärvi and lake management planning; validation of the algorithms of the GOMOS ozone remote sensing instrument on board the Envisat satellite of European Space Agency and the study of the effects of aerosol model selection on the GOMOS algorithm.

Assessment of Coastal Bird Populations and Habitats on the Finnish Coast of the Baltic Sea: Implications for Monitoring and Management

Coastal birds are an integral part of coastal ecosystems, which nowadays are subject to severe environmental pressures. Effective measures for the management and conservation of seabirds and their habitats call for insight into their population processes and the factors affecting their distribution and abundance. Central to national and international management and conservation measures is the availability of accurate data and information on bird populations, as well as on environmental trends and on measures taken to solve environmental problems. In this thesis I address different aspects of the occurrence, abundance, population trends and breeding success of waterbirds breeding on the Finnish coast of the Baltic Sea, and discuss the implications of the results for seabird monitoring, management and conservation. In addition, I assess the position and prospects of coastal bird monitoring data, in the processing and dissemination of biodiversity data and information in accordance with the Convention on Biological Diversity (CBD) and other national and international commitments. I show that important factors for seabird habitat selection are island area and elevation, water depth, shore openness, and the composition of island cover habitats. Habitat preferences are species-specific, with certain similarities within species groups. The occurrence of the colonial Arctic Tern (Sterna paradisaea) is partly affected by different habitat characteristics than its abundance. Using long-term bird monitoring data, I show that eutrophication and winter severity have reduced the populations of several Finnish seabird species. A major demographic factor through which environmental changes influence bird populations is breeding success. Breeding success can function as a more rapid indicator of sublethal environmental impacts than population trends, particularly for long-lived and slowbreeding species, and should therefore be included in coastal bird monitoring schemes. Among my target species, local breeding success can be shown to affect the populations of the Mallard (Anas platyrhynchos), the Eider (Somateria mollissima) and the Goosander (Mergus merganser) after a time lag corresponding to their species-specific recruitment age. For some of the target species, the number of individuals in late summer can be used as an easier and more cost-effective indicator of breeding success than brood counts. My results highlight that the interpretation and application of habitat and population studies require solid background knowledge of the ecology of the target species. In addition, the special characteristics of coastal birds, their habitats, and coastal bird monitoring data have to be considered in the assessment of their distribution and population trends. According to the results, the relationships between the occurrence, abundance and population trends of coastal birds and environmental factors can be quantitatively assessed using multivariate modelling and model selection. Spatial data sets widely available in Finland can be utilised in the calculation of several variables that are relevant to the habitat selection of Finnish coastal species. Concerning some habitat characteristics field work is still required, due to a lack of remotely sensed data or the low resolution of readily available data in relation to the fine scale of the habitat patches in the archipelago. While long-term data sets exist for water quality and weather, the lack of data concerning for instance the food resources of birds hampers more detailed studies of environmental effects on bird populations. Intensive studies of coastal bird species in different archipelago areas should be encouraged. The provision and free delivery of high-quality coastal data concerning bird populations and their habitats would greatly increase the capability of ecological modelling, as well as the management and conservation of coastal environments and communities. International initiatives that promote open spatial data infrastructures and sharing are therefore highly regarded. To function effectively, international information networks, such as the biodiversity Clearing House Mechanism (CHM) under the CBD, need to be rooted at regional and local levels. Attention should also be paid to the processing of data for higher levels of the information hierarchy, so that data are synthesized and developed into high-quality knowledge applicable to management and conservation.

Kernel-Based Ranking. Methods for Learning and Performance Estimation

Machine learning provides tools for automated construction of predictive models in data intensive areas of engineering and science. The family of regularized kernel methods have in the recent years become one of the mainstream approaches to machine learning, due to a number of advantages the methods share. The approach provides theoretically well-founded solutions to the problems of under- and overfitting, allows learning from structured data, and has been empirically demonstrated to yield high predictive performance on a wide range of application domains. Historically, the problems of classification and regression have gained the majority of attention in the field. In this thesis we focus on another type of learning problem, that of learning to rank. In learning to rank, the aim is from a set of past observations to learn a ranking function that can order new objects according to how well they match some underlying criterion of goodness. As an important special case of the setting, we can recover the bipartite ranking problem, corresponding to maximizing the area under the ROC curve (AUC) in binary classification. Ranking applications appear in a large variety of settings, examples encountered in this thesis include document retrieval in web search, recommender systems, information extraction and automated parsing of natural language. We consider the pairwise approach to learning to rank, where ranking models are learned by minimizing the expected probability of ranking any two randomly drawn test examples incorrectly. The development of computationally efficient kernel methods, based on this approach, has in the past proven to be challenging. Moreover, it is not clear what techniques for estimating the predictive performance of learned models are the most reliable in the ranking setting, and how the techniques can be implemented efficiently. The contributions of this thesis are as follows. First, we develop RankRLS, a computationally efficient kernel method for learning to rank, that is based on minimizing a regularized pairwise least-squares loss. In addition to training methods, we introduce a variety of algorithms for tasks such as model selection, multi-output learning, and cross-validation, based on computational shortcuts from matrix algebra. Second, we improve the fastest known training method for the linear version of the RankSVM algorithm, which is one of the most well established methods for learning to rank. Third, we study the combination of the empirical kernel map and reduced set approximation, which allows the large-scale training of kernel machines using linear solvers, and propose computationally efficient solutions to cross-validation when using the approach. Next, we explore the problem of reliable cross-validation when using AUC as a performance criterion, through an extensive simulation study. We demonstrate that the proposed leave-pair-out cross-validation approach leads to more reliable performance estimation than commonly used alternative approaches. Finally, we present a case study on applying machine learning to information extraction from biomedical literature, which combines several of the approaches considered in the thesis. The thesis is divided into two parts. Part I provides the background for the research work and summarizes the most central results, Part II consists of the five original research articles that are the main contribution of this thesis.

Computation and Analysis of Multiple Structural-Change Models

In a recent paper, Bai and Perron (1998) considered theoretical issues related to the limiting distribution of estimators and test statistics in the linear model with multiple structural changes. In this companion paper, we consider practical issues for the empirical applications of the procedures. We first address the problem of estimation of the break dates and present an efficient algorithm to obtain global minimizers of the sum of squared residuals. This algorithm is based on the principle of dynamic programming and requires at most least-squares operations of order O(T 2) for any number of breaks. Our method can be applied to both pure and partial structural-change models. Secondly, we consider the problem of forming confidence intervals for the break dates under various hypotheses about the structure of the data and the errors across segments. Third, we address the issue of testing for structural changes under very general conditions on the data and the errors. Fourth, we address the issue of estimating the number of breaks. We present simulation results pertaining to the behavior of the estimators and tests in finite samples. Finally, a few empirical applications are presented to illustrate the usefulness of the procedures. All methods discussed are implemented in a GAUSS program available upon request for non-profit academic use.

Sélection de modèle d'imputation à partir de modèles bayésiens hiérarchiques linéaires multivariés

Les logiciels utilisés sont Splus et R.

Distributions d'auto-amorçage exactes ponctuelles des courbes ROC et des courbes de coûts

Mémoire numérisé par la Division de la gestion de documents et des archives de l'Université de Montréal

Choix de portefeuille de grande taille et mesures de risque pour preneurs de décision pessimistes

Cette thèse de doctorat consiste en trois chapitres qui traitent des sujets de choix de portefeuilles de grande taille, et de mesure de risque. Le premier chapitre traite du problème d’erreur d’estimation dans les portefeuilles de grande taille, et utilise le cadre d'analyse moyenne-variance. Le second chapitre explore l'importance du risque de devise pour les portefeuilles d'actifs domestiques, et étudie les liens entre la stabilité des poids de portefeuille de grande taille et le risque de devise. Pour finir, sous l'hypothèse que le preneur de décision est pessimiste, le troisième chapitre dérive la prime de risque, une mesure du pessimisme, et propose une méthodologie pour estimer les mesures dérivées. Le premier chapitre améliore le choix optimal de portefeuille dans le cadre du principe moyenne-variance de Markowitz (1952). Ceci est motivé par les résultats très décevants obtenus, lorsque la moyenne et la variance sont remplacées par leurs estimations empiriques. Ce problème est amplifié lorsque le nombre d’actifs est grand et que la matrice de covariance empirique est singulière ou presque singulière. Dans ce chapitre, nous examinons quatre techniques de régularisation pour stabiliser l’inverse de la matrice de covariance: le ridge, spectral cut-off, Landweber-Fridman et LARS Lasso. Ces méthodes font chacune intervenir un paramètre d’ajustement, qui doit être sélectionné. La contribution principale de cette partie, est de dériver une méthode basée uniquement sur les données pour sélectionner le paramètre de régularisation de manière optimale, i.e. pour minimiser la perte espérée d’utilité. Précisément, un critère de validation croisée qui prend une même forme pour les quatre méthodes de régularisation est dérivé. Les règles régularisées obtenues sont alors comparées à la règle utilisant directement les données et à la stratégie naïve 1/N, selon leur perte espérée d’utilité et leur ratio de Sharpe. Ces performances sont mesurée dans l’échantillon (in-sample) et hors-échantillon (out-of-sample) en considérant différentes tailles d’échantillon et nombre d’actifs. Des simulations et de l’illustration empirique menées, il ressort principalement que la régularisation de la matrice de covariance améliore de manière significative la règle de Markowitz basée sur les données, et donne de meilleurs résultats que le portefeuille naïf, surtout dans les cas le problème d’erreur d’estimation est très sévère. Dans le second chapitre, nous investiguons dans quelle mesure, les portefeuilles optimaux et stables d'actifs domestiques, peuvent réduire ou éliminer le risque de devise. Pour cela nous utilisons des rendements mensuelles de 48 industries américaines, au cours de la période 1976-2008. Pour résoudre les problèmes d'instabilité inhérents aux portefeuilles de grandes tailles, nous adoptons la méthode de régularisation spectral cut-off. Ceci aboutit à une famille de portefeuilles optimaux et stables, en permettant aux investisseurs de choisir différents pourcentages des composantes principales (ou dégrées de stabilité). Nos tests empiriques sont basés sur un modèle International d'évaluation d'actifs financiers (IAPM). Dans ce modèle, le risque de devise est décomposé en deux facteurs représentant les devises des pays industrialisés d'une part, et celles des pays émergents d'autres part. Nos résultats indiquent que le risque de devise est primé et varie à travers le temps pour les portefeuilles stables de risque minimum. De plus ces stratégies conduisent à une réduction significative de l'exposition au risque de change, tandis que la contribution de la prime risque de change reste en moyenne inchangée. Les poids de portefeuille optimaux sont une alternative aux poids de capitalisation boursière. Par conséquent ce chapitre complète la littérature selon laquelle la prime de risque est importante au niveau de l'industrie et au niveau national dans la plupart des pays. Dans le dernier chapitre, nous dérivons une mesure de la prime de risque pour des préférences dépendent du rang et proposons une mesure du degré de pessimisme, étant donné une fonction de distorsion. Les mesures introduites généralisent la mesure de prime de risque dérivée dans le cadre de la théorie de l'utilité espérée, qui est fréquemment violée aussi bien dans des situations expérimentales que dans des situations réelles. Dans la grande famille des préférences considérées, une attention particulière est accordée à la CVaR (valeur à risque conditionnelle). Cette dernière mesure de risque est de plus en plus utilisée pour la construction de portefeuilles et est préconisée pour compléter la VaR (valeur à risque) utilisée depuis 1996 par le comité de Bâle. De plus, nous fournissons le cadre statistique nécessaire pour faire de l’inférence sur les mesures proposées. Pour finir, les propriétés des estimateurs proposés sont évaluées à travers une étude Monte-Carlo, et une illustration empirique en utilisant les rendements journaliers du marché boursier américain sur de la période 2000-2011.

Comparing Support Vector Machines with Gaussian Kernels to Radial Basis Function Classifiers

The Support Vector (SV) machine is a novel type of learning machine, based on statistical learning theory, which contains polynomial classifiers, neural networks, and radial basis function (RBF) networks as special cases. In the RBF case, the SV algorithm automatically determines centers, weights and threshold such as to minimize an upper bound on the expected test error. The present study is devoted to an experimental comparison of these machines with a classical approach, where the centers are determined by $k$--means clustering and the weights are found using error backpropagation. We consider three machines, namely a classical RBF machine, an SV machine with Gaussian kernel, and a hybrid system with the centers determined by the SV method and the weights trained by error backpropagation. Our results show that on the US postal service database of handwritten digits, the SV machine achieves the highest test accuracy, followed by the hybrid approach. The SV approach is thus not only theoretically well--founded, but also superior in a practical application.