ABSTRACT
In this paper we examine the accuracy and precision of three indices of catch-per-unit-effort (CPUE). We carried out simulations, generating catch data according to six probability distributions (normal, Poisson, lognormal, gamma, delta and negative binomial), three variance structures (constant, proportional to effort and proportional to the squared effort) and their magnitudes (tail weight). The Jackknife approach of the index is recommended, whenever catch is proportional to effort or even under small deviations from proportionality assumption, when a ratio estimator is to be applied and little is known about the underlying behaviour of variables, as is the case for most fishery studies.
Neste trabalho, examinamos a acurácia e precisão de três índices de captura por unidade de esforço (CPUE). Foram feitas simulações, nas quais foram gerados dados de captura de acordo com seis distribuições de probabilidade (normal, Poisson, lognormal, gama, delta e binomial negativa), três estruturas de variância (constante, proporcional ao esforço e proporcional ao quadrado do esforço), e magnitudes (tail weight). É recomendado o uso do método Jackknife para os índices, sempre que a captura for proporcional ao esforço ou até em casos de pequenos desvios do pressuposto de proporcionalidade, quando se deseja utilizar um estimador de razão e pouco é conhecido sobre o real comportamento das variáveis, como é o caso da maioria dos estudos de pesca.
ABSTRACT
In this paper we examine the accuracy and precision of three indices of catch-per-unit-effort (CPUE). We carried out simulations, generating catch data according to six probability distributions (normal, Poisson, lognormal, gamma, delta and negative binomial), three variance structures (constant, proportional to effort and proportional to the squared effort) and their magnitudes (tail weight). The Jackknife approach of the index is recommended, whenever catch is proportional to effort or even under small deviations from proportionality assumption, when a ratio estimator is to be applied and little is known about the underlying behaviour of variables, as is the case for most fishery studies.
Neste trabalho, examinamos a acurácia e precisão de três índices de captura por unidade de esforço (CPUE). Foram feitas simulações, nas quais foram gerados dados de captura de acordo com seis distribuições de probabilidade (normal, Poisson, lognormal, gama, delta e binomial negativa), três estruturas de variância (constante, proporcional ao esforço e proporcional ao quadrado do esforço), e magnitudes (tail weight). É recomendado o uso do método Jackknife para os índices, sempre que a captura for proporcional ao esforço ou até em casos de pequenos desvios do pressuposto de proporcionalidade, quando se deseja utilizar um estimador de razão e pouco é conhecido sobre o real comportamento das variáveis, como é o caso da maioria dos estudos de pesca.
Subject(s)
Animals , Biometry , Fisheries/statistics & numerical data , Models, Statistical , Monte Carlo Method , ProbabilityABSTRACT
It is hypothesized that the number of outpatient visits can be represented by three different probability models: the truncated Poisson distribution, the Zeta distribution and the logarithmic series distribution. Maximum likelihood estimates of parameters of the above distributions were obtained by using grouped data according to the number of visits. A goodness-of-fit test was also made to compare the fit of the three distributions and the value of this statistic was classified and compared according to the types of medical care facilities. Based on the likelihood ratio statistic as a test criterion, both the truncated Poisson and Zeta distributions were not appropriate for the model of the number of outpatient visits. However, the logarithmic series distribution provides a good fit to data in the case of university hospitals, general hospitals and hospitals. When we apply this distribution in the 10 most common diseases, the estimates of the parameter vary from 0.39567 to 0.54176 for university hospitals, from 0.45329 to 0.65387 for general hospitals, and from 0.55104 to 0.77625 for hospitals. On the other hand in the case of clinics, even the logarithmic series distribution cannot be fitted to the data well. A characteristic of clinic utilization with almost homogeneous treatment patterns, in spite of the fact that there are a great many clinics, could be the reason for the above results.