TY - GEN
T1 - Evaluating goodness-of-fit in comparison of different expressions for length-weight relationship in fishery resources
AU - Xu, Hai Long
AU - Chen, Yong
AU - Gu, De Xian
AU - Qiao, Xiu Ting
N1 - Publisher Copyright:
© (2014) Trans Tech Publications, Switzerland.
PY - 2014
Y1 - 2014
N2 - To find out the optimal length-weight relationship expression of fishery species, eight different expressions were used to fit the length-weight relationships in 18 fishery species (nine fish species, two shrimp species, three cephalopoda species, three crab species and one mantis shrimp species), including Linear Regression Function, Body Mass Index, Power Regression Function and Polynomial Function, and the goodness-of-fit of each expressions were compared by adjusted R-square, Residual Standard Deviation(RSD), and Fitting Optimization Index of Curve Regression which is proposed by Zhang. The results turned out that the polynomial equation W = a + b * L3 + c * L2 + d * L showed the best goodness-of-fit, while the linear equation W = a + b * L and BMI equation W = a * L2 showed the worst goodness-of-fit. Combined with prediction curve of regression function, there may be over-fitting in the model of polynomial function. In this research, the expressions W = a + b * Lc was proved to show the best goodness-of-fit in length-weight relationship at given simples or populations, and the constant term ‘a’ as an error may not be zero for samples.
AB - To find out the optimal length-weight relationship expression of fishery species, eight different expressions were used to fit the length-weight relationships in 18 fishery species (nine fish species, two shrimp species, three cephalopoda species, three crab species and one mantis shrimp species), including Linear Regression Function, Body Mass Index, Power Regression Function and Polynomial Function, and the goodness-of-fit of each expressions were compared by adjusted R-square, Residual Standard Deviation(RSD), and Fitting Optimization Index of Curve Regression which is proposed by Zhang. The results turned out that the polynomial equation W = a + b * L3 + c * L2 + d * L showed the best goodness-of-fit, while the linear equation W = a + b * L and BMI equation W = a * L2 showed the worst goodness-of-fit. Combined with prediction curve of regression function, there may be over-fitting in the model of polynomial function. In this research, the expressions W = a + b * Lc was proved to show the best goodness-of-fit in length-weight relationship at given simples or populations, and the constant term ‘a’ as an error may not be zero for samples.
KW - Comparison
KW - Goodness-of-fit
KW - Length-weight relationship
UR - https://www.scopus.com/pages/publications/84914669269
U2 - 10.4028/www.scientific.net/AMM.651-653.337
DO - 10.4028/www.scientific.net/AMM.651-653.337
M3 - Conference contribution
AN - SCOPUS:84914669269
T3 - Applied Mechanics and Materials
SP - 337
EP - 343
BT - Material Science, Civil Engineering and Architecture Science, Mechanical Engineering and Manufacturing Technology II
A2 - Liu, H.W.
A2 - Wang, G.
A2 - Zhang, G.W.
PB - Trans Tech Publications Ltd
T2 - 3rd International Conference on Advanced Engineering Materials and Architecture Science, ICAEMAS 2014
Y2 - 26 July 2014 through 27 July 2014
ER -