Hello dear researchers, how can I determine the constant "b" from the application of the exponential regression equation W = aL ^ b (Ricker, 1975) such that "a" and "b" are constants and length: (L) and weight: (w) in STATISTICA SOFTWAR.
Hello Youssef,
You have to put your equation in logarithmic form to get a linear function of type
y = ax + b.
W = aL ^ b
Log (W) = Log (a) + bLog (L).
You then determine a and b from Log (a) and Log (b).
Good luck.
Hello Youssef,
You have to put your equation in logarithmic form to get a linear function of type
y = ax + b.
W = aL ^ b
Log (W) = Log (a) + bLog (L).
You then determine a and b from Log (a) and Log (b).
Good luck.
Hello Youssef,
You can select parameter values from previous studies or choose their values from a solid article.
Thank you very much Dears researchers (SSI Abderrahmane Zekhnini, Ali S. Habeeb and Ette Etuk) for your answers.
You can take log of a common base on both sides of the equation.
Then apply the multiplication laws and power law of log.
El Mourabit Youssef, If you are looking for a least-squares solution, logarithmic transformations are not the right way. Please look at https://www.researchgate.net/publication/304232132_Transformaciones_logaritmicas_en_regresion_simple
where I show that this procedure can lead to very different solutions from those of least squares.
Thank you Dear Sir Jorge Ortiz Pinilla for your answer, I will try it.
You can get the estimates of a and b in two ways:
1) run the analysis as ordinary nonlinear least square (ONLS), i.e., nonlinear regression. You will have to specify initial values for the parameters. Setting a =1 & b=1 should be good starting values for the parameters in this model. Bit if convergence is not achieved, you can increase the initial value of a.
2) the second option is to do a logarithm transformation and run the analysis as ordinary least square (OLS), i.e., linear regression as recommended by other experts. However, a back transformation would be required. A correction factor (CF) to account for back transformation error must be computed. CF = exp(MSE/2), where MSE represents the mean square error.
The parameters of exponential regression equation can be found by SigmaPlot program
I do not know which software you are using; excel, STATISICA, SPSS, R, SAS or anything else. It is simple non linear equation. You tabulate values of W and L in excel, use non linear regression tool from any of these software, you will get the parameter estimate for a and b. In software like SAS, you need to assign some initial starting value for a and b. It you find difficulty in judging initial value for a and b, see any publication using similar equation and use those values in your work as initial value or put any minimum value considering the dataset you have.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression