Premium
Stock Differentiation of Walleye Based on the Fourier Approximation of Averaged Scale Outline Signals
Author(s) -
Watkinson Douglas A.,
Gillis Darren M.
Publication year - 2003
Publication title -
north american journal of fisheries management
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.587
H-Index - 72
eISSN - 1548-8675
pISSN - 0275-5947
DOI - 10.1577/1548-8675(2003)023<0091:sdowbo>2.0.co;2
Subject(s) - linear discriminant analysis , nonparametric statistics , discriminant , statistics , econometrics , mathematics , statistical hypothesis testing , scale (ratio) , stock (firearms) , fourier transform , discriminant function analysis , multivariate statistics , jackknife resampling , pattern recognition (psychology) , artificial intelligence , computer science , geography , cartography , mathematical analysis , archaeology , estimator
Fourier analysis of scale outlines has been successfully used as a first method of stock discrimination in a number of fish species. However, the refinement of methods that increase the ability to differentiate among stocks would improve its utility for fisheries management. Signals created from averaging the scales of several walleyes Stizostedion vitreum produced variables that formed significantly better discriminant functions for stock discrimination than the single‐scale signals traditionally used due to the variability of scale outline shapes from individual fish. Fourier analysis was used to quantify scale outline shape; the variables produced violated the discriminant analysis assumption of multivariate normality. A nonparametric randomization procedure, new to stock discrimination studies, was developed to test the significance of the discriminant functions formed. All signal‐comparison discriminant functions were significant. A modified jackknife procedure was also developed to test for significant differences between the discriminant functions formed from the single‐scale and averaged‐scale signal variables. Based on our findings, multiple‐scale signals should be used in future stock discrimination studies. The nonparametric statistics used here can test for significance when the assumptions of classical discriminant analysis are violated.