B1011
Title: Modelling independent and preferential data jointly
Authors: Mario Figueira Pereira - Universitat de Valencia (Spain) [presenting]
David Conesa - Universitat de Valencia (Spain)
Antonio Lopez-Quilez - University of Valencia (Spain)
Iosu Paradinas - AZTI (Spain)
Abstract: Species distribution models (SDMs) in continuous space have been extensively used as a valuable tool in ecological statistical analysis. In ecology, two common models employed are geostatistical models and preferential models. Geostatistical models are suitable when the process being studied is independent of the sampling locations, whereas preferential models are appropriate when the sampling locations depend on the process under study. However, what if both types of data collected for the same process are obtained? The aim is to explore the suitability of geostatistical models, preferential models, and a mixture model that accounts for the different sampling schemes. The results indicate that, in general, the preferential and mixture models yield satisfactory and closely aligned results in most cases. On the other hand, geostatistical models consistently produce inferior estimates when faced with higher spatial complexity, a smaller number of samples, and a lower proportion of completely random samples.
