Dalhousie Statistics Seminars 2017/2018

If you would like to schedule a talk or be added to, or removed from, the mailing list, please email Edward Susko. Here is the schedule:

Abstracts

Abstract: Molecular sequence data that have evolved under the influence of heterotachous evolutionary processes are known to mislead phylogenetic inference. We introduce the General Heterogeneous evolution On a Single Topology (GHOST) model of sequence evolution, implemented under a maximum-likelihood framework in the phylogenetic program IQ-TREE (http://www.iqtree.org). Our simulations show that using the GHOST model, IQ-TREE can accurately recover the tree topology, branch lengths and substitution model parameters from heterotachously-evolved sequences. We apply our model to an electric fish dataset and identify a subtle component of the historical signal, linked to the previously established convergent evolution of the electric organ in two geographically distinct lineages of electric fish. We compare inference under the GHOST model to partitioning by codon position and show that, owing to the minimization of model constraints, the GHOST model is able to offer unique biological insights when applied to empirical data.

Montse Fuentes (Department of Statistical Sciences and Operations Research and of Biostatistics, Virginia Commonwealth University): "Nonparametric Spatial-Temporal Modelling of the Association between Ambient Air Pollution and Adverse Pregnancy Outcomes"

Abstract: Exposure to high levels of air pollution during the pregnancy is associated with increased probability of birth defects, a major cause of infant morbidity and mortality. New statistical methodology is required to specifically determine when a particular pollutant impacts the pregnancy outcome, to determine the role of different pollutants, and to characterize the spatial variability in these results.

We introduce a new methodology for high dimensional environmental-health data. More specifically, we present a Bayesian spatial-temporal hierarchical multivariate probit regression model that identifies weeks during the first trimester of pregnancy which are impactful in terms of cardiac congenital anomaly development. The model is able to consider multiple pollutants and a multivariate cardiac anomaly grouping outcome jointly while allowing the critical windows to vary in a continuous manner across time and space. We utilize a dataset of numerical chemical model output which contains information regarding multiple species of particulate matter. Our introduction of an innovative spatial-temporal nonparametric prior distribution for the pollution risk effects allows for greater flexibility to identify critical weeks during pregnancy which are missed when more standard models are applied. We apply these methods to geocoded pregnancy outcomes in Texas.

Edward Susko (Department of Mathematics and Statistics, Dalhousie University): "Bayes Factor Biases for Non-Nested Models and Corrections"

Abstract: With the advent of simulation-based methods to obtain samples from posteriors and due to increases in computational power, Bayesian methods are increasingly applied to complex problems, sometimes providing the only available methods where likelihood implementations are difficult. As a consequence, a large body of research in science and social science increasingly utilizes Bayesian tools, often applying them with default settings. A fundamental problem of interest is model selection and Bayes factors provide a natural approach to Bayesian model selection. Using Laplace approximations and illustrative examples, we demonstrate that Bayes factors can have strong biases towards particular models even in non-nested settings with the same number of parameters. Several easily implemented corrections are shown to provide effective cross-checks to default Bayes Factors.

Paul McNicholas (Department of Mathematics and Statistics, McMaster University): "Clustering Via Mixture Models"

Abstract: The application of mixture models for clustering has grown into an important subfield of classification. First, the framework for mixture model-based clustering is established and some historical context is provided. Then, some previous work is reviewed before more recent work is presented. This includes work on clustering in the presence of outlying points as well as approaches for asymmetric clusters and high-dimensional data. The talk concludes with a discussion about ongoing and future work, including some work on mixtures of matrix variate distributions.