It Is convenient to use the coefficients derived from performing the 

 linear regression on log-transformed data as the Initial estimates (a^* 

 bg) for the Iteration process. If we define a new function In terms of 

 estimated coefficients as, 



b 

 y* - f<x,aQ,bo) - a^x ° 



the discrete values of this approximating function will be, 



y'l - ^(^i*^o»\y 



y'2 - f(x2,ao,bo) 



y'n ■ f(xn»ao»^o^ 



where n Is the total number of data pairs used In the regression 

 equations. 



Realizing that the coefficients a and b produce the "best fit" solu- 

 tions, the minimized residuals are represented as, 



Vj - f(xj,a,b) - 71 

 Vj - f(x2,a,b) - 72 



^n " f(xn,a,b) - y„ 



where y^, y2>*««yQ are the observed dependent variables. Substituting 

 our approximation yields 



127 



