MACSI at the department of Mathematics and Statistics at the University of Limerick invites you to a seminar
Date: Friday, 1st February 2019
Speaker: Cóilín Minto (Marine and Freshwater Research Centre, GMIT, Galway. Work conducted in collaboration with many colleagues at various institutions to be acknowledged).
Title: Challenges of missing data in fisheries science.
Abstract: Statistical modelling is central to fisheries science with applications ranging from understanding individual life histories, population status and community dynamics. Outputs from these models are often used as the basis for resource management advice. Challenges abound, however, including pervasive missing data. Missingness is application dependent, here I focus on two disparate settings where missing data challenges inference.
Developing from previous work on growth modelling, a first application concerns estimating sex-specific maturity models. Sex-specific maturity models are typically estimated by fitting logit-linear curves with a binomial distributional assumption to individuals of known sex. Yet, macrospically ascribing sex can be difficult, particularly for immature animals. As a result, sex-specific maturity models are often fit to known-sex individuals, omitting unclassified immature individuals. Worse, when fitting maturity models juvenile animals are often used twice, once for female and once for male curve estimation. This results in biased and uncertain parameter estimates. An alternative approach is proposed where the sex of the unclassified individuals is treated as a missing data problem. Analytical and EM-algorithm solutions will be compared.
A second application concerns monitoring the performance of stated policy goals (e.g., EU Common Fisheries Policy). Overall performance often requires summarising the status and pressures across multiple populations in a given region. Time series start and end at different time points resulting in sensitivity of raw summary statistics to the set of series present in given years. A hierarchical solution is proposed, whereby subject-specific curves are allowed to deviate from the global trend in a variety of ways including first- and second-order penalised deviations and more typical time series deviations. Influence of these assumptions on the global trend will be explored.
Supported by Science Foundation Ireland funding, MACSI - the Mathematics Applications Consortium for Science and Industry (www.macsi.ul.ie), centred at the University of Limerick, is dedicated to the mathematical modelling and solution of problems which arise in science, engineering and industry in Ireland.