Title: Generalized linear latent variable models for analyzing multivariate abundance data
Authors: Sara Taskinen - University of Jyvaskyla (Finland) [presenting]
Jenni Niku - University of Jyvaskyla (Finland)
Francis Hui - Australian National University (Australia)
David Warton - University of New South Wales (Australia)
Abstract: In ecological studies, abundances of many, interacting species are often collected in several sites. Such data are often very sparse, high-dimensional and include highly correlated responses. The main aim of the statistical analysis is then to understand the interrelationships among such multiple, correlated responses. We consider model-based approaches for analyzing multivariate abundance data. We will show how generalized linear latent variable models (GLLVMs) can easily capture the correlation inherent in responses and provide a powerful tool for estimation and inference. Fast and efficient maximum likelihood based algorithms for fitting the models will be discussed. It is shown that especially variational approximation method performs better than several classical estimation methods for GLLVMs. The methods will be applied to ecological datasets focusing on model-based approaches to unconstrained ordination.