B1513
Title: Analysing recreational boating traffic data in a panel time series data modelling setting
Authors: Ebenezer Afrifa-Yamoah - Edith Cowan University (Australia) [presenting]
Stephen M Taylor - Department of Primary Industries and Regional Development (Austria)
Ute Mueller - Edith Cowan University (Australia)
Abstract: The lack of continuity in recreational fisheries data due to intermittent sampling or surveys make trend estimation of effort difficult. Using a pooled time-series cross-sectional study design, panel generalized linear modelling techniques were used to identify potential determinants of recreational boating effort. Panels of data comprised time-lapse camera monitoring data of recreational boating effort and climatic and calendar-based variables over the period of 2011-2016 for four locations in different bioregions of Western Australia. Long-term equilibrium effects of the predictors on recreational boating traffic were estimated for within, between, random and pooled effects Poisson models. Significant effects ($p < 0.001$) on recreational boating traffic were observed for temperature, wind speed and direction, precipitation, sea level pressure, humidity, time of day and month of the year across all panels. Sub-panel analyses revealed varying levels of importance of predictors with respect to locations. Non-linear tests of causality using artificial neural networks (ANN) based on vector auto-regressive neural network established significant unidirectional Granger causality of the study predictors on the recreational boating traffic.
