UW Aquatic & Fishery Sciences Quantitative Seminar
Hiroshi Okamura
National Research Institute of Far Seas Fisheries
Statistical analysis for temporal periodicity in fish age validation
Abstract
Information on the ages of fish populations is fundamental to the estimation of biological parameters and accurate estimation of age is therefore essential to successful fishery management. I focus on the age validation, which is to estimate the periodicity of growth increment formation. Although various approaches exist for validating the age estimation of fish, I deal with edge analysis (EA) and marginal increment analysis (MIA) because they are the most commonly used methods. EA and MIA are frequently based on visual assessments or a simple analysis of variance is conducted at best. However, temporal periodicity can invalidate traditional methods. I present new statistical methods for EA and MIA taking a feature of circular data into account where basic modeling is different between EA and MIA according to specific characteristics of their data. The periodicity of growth band pairs is categorized as no cycle, an annual cycle, or a biannual cycle. Three models are then constructed according to different periodicities and the mixture distributions are effectively used for both of EA and MIA. We use the Akaike information criterion (AIC) to determine the best model. The general performance of the methods was evaluated using simulated data of various sample sizes. The methods should improve the accuracy of age determination and could be applied to all species that have periodic growth band pairs.