POSTERIOR DE POLYA NO MONITORAMENTO AMOSTRAL DE PESCARIAS

Abstract

A non-informative bayesian approach to sample-based fishery surveys is proposed. The Polya posterior for finite population parameters is used to obtain the inferences. The viability of a sampling plan was used in a pilot field experiment to collect weekly information about effort and catch from the artisanal fishery in Rio Grande, RS. Based on a simulated virtual population with four species and 345 fishermen, the sampling plan was tested using a sampling fraction of 3.3% from a complete data matrix of 2760 components. Results have shown accuracies above 71% for all but the most problematic species 2, and around 90% for estimates of total catch and cummulative effort. The percentile probability intervals (ICr) perform slightly better than the highest density interval (HDI) in terms of coverage; although both resulted about 5 percentage points bellow the nominal value of 95%.