A comparison of weighted least square estimation and ordinary least square estimation for analysing weight-length relationship of Por’s Goatfish (Upeneus pori Ben-Tuvia & Golani, 1989) in Iskenderun bay

Abstract

A weighted least square (WLS) and ordinary least square (OLS) approach for analysing fish length-weight relationships, was applied. We reviewed and compared two regression methods used in statistics to reconstruct length-weight relationship from Por’s Goatfish (Upeneus pori Ben-Tuvia & Golani, 1989) length-weight data in Iskenderun Bay. This comparison relies on a suite of regression model verification statistics to validate the accuracy of the length-weight relationship produced by the best model. The results indicate WLS may perform slightly better in heteroscedastic data; conversely, OLS may perform slightly better when the data is homoscedastic.