For shape such as y=x1 x2 ^ ^ b2 * b3 b1 * * x3 ^ b4 function, when using multivariate nonlinear regression nlinfit function, need to set parameters of the initial value of b, and the initial values have an effect on the final result, but take the log on both sides of the function, can get: the log (y)=log (b1) + b2 * log (x1) + b3 * log (x2) + b4 * log (x3), regress function can be used for fitting at this time,

Nlinfit function fitting result set to Y1, regress function fitting is set to log (Y2), again to Y2, by comparison, the result of the Y1 and Y2 not equal (high), do you know the problem?