Regress nonlinear equation

Originally Posted by rwgrietveld

what is Ln(0) How can C=0 (zero) give a result? it is -INF

Yes, I realize that the ln(0) is indeterminate. Perhaps one can use a very small number instead, like 0.00001 as an example.

Originally Posted by shg

And why the prejudice against Solver?

I'm trying to match ther output from an analytical instrument that is "ancient", actually at least 20 years old. Obviously, it doesn't use Solver. There's got to be an old-fashioned step-by-step way to determine these parameters. Basically I'm trying to find out what algorithm is being used in the instrument.

I'm familiar with Newton-Raphson, but only for one variable, not 4 or 5. Certainly there must be a way in Excel to do these calculations the "hard" way.

Originally Posted by shg

It regresses well using a small value for C in lieu of zero. And of course it's possible to do (eventually) via VBA or even worksheet functions what Solver does.

Two things: First, it might be hard to charm someone into reinventing Solver. Second, your answer makes it sound like you would further wish to iterate over several methods of solution to identify some 'best fit' to instrument results. Tall order, methinks.

Why not try Solver first?

Out of curiosity, what instrument?

CBAS MIRA This is a clinical chemistry random access analyzer. It measures components in a person's blood serum or plasma

I don't know what you got for results but the instrument output is the following:
Ro = 486.352
Kc = 839.297
a = -4.87399
b = 1.99643

I'll bet you didn't get these numbers.

I can't use Solver. There's got to be a way. This instrument doesn't have Solver.

Originally Posted by shg

Not, I get nothing like that -- but those numbers don't remotely fit the model, either (see attached).

As a general comment, I think a regression should have at least three times as many data points as coefficients; this has four data points and four coefficients.

Actually the data does fit the parameters (see attached) If you solve for C using the calculated R values, along with the 4 parameters Ro, Kc, a and b it will return the values for each standard.

Regardless, Istill need to find out the algorithm that detrmines th parms that the instrument spits out.

Thanks