Premium
A New Method to Compute Standard‐Weight Equations That Reduces Length‐Related Bias
Author(s) -
Gerow Kenneth G.,
AndersonSprecher Richard C.,
Hubert Wayne A.
Publication year - 2005
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/m04-196.1
Subject(s) - percentile , computation , curvilinear coordinates , mathematics , simultaneous equations , regression , regression analysis , line (geometry) , statistics , algorithm , mathematical analysis , geometry , differential equation
We propose a new method for developing standard‐weight ( W s ) equations for use in the computation of relative weight ( W r ) because the regression line–percentile (RLP) method often leads to length‐related biases in W s equations. We studied the structural properties of W s equations developed by the RLP method through simulations, identified reasons for biases, and compared W s equations computed by the RLP method and the new method. The new method is similar to the RLP method but is based on means of measured weights rather than on means of weights predicted from regression models. The new method also models curvilinear W s relationships not accounted for by the RLP method. For some length‐classes in some species, the relative weights computed from W s equations developed by the new method were more than 20 W r units different from those using W s equations developed by the RLP method. We recommend assessment of published W s equations developed by the RLP method for length‐related bias and use of the new method for computing new W s equations when bias is identified.