Re: non linear regression
Friday, May 11, 2007 10:09 PM
I was looking in one of my old texts about this sort of thing. Seems like exponentials are generally linearized (when possible) by working in log-space and re-expressing the equation of interest. This [sometimes] permits treating the question in a linear fashion: ln(F) = ln(exp^(a*x+b) + exp^(c*x+d)) ... but I didn't get far beyond blinking at that intial step and putting the book away. (Haven't had to use calculus since leaving grad school over ten years ago, so my brain doesn't remember the rules on expanding log expressions to go any further.) That probably isn't the kind of information you are wanting, though.