[proFit-list] statistically valid fit?
Bruce Palfey
brupalf at umich.edu
Wed Jan 2 19:38:49 CST 2008
I have data from kinetics experiments that I'd like to fit. The
differential equations describing the mechanism can be integrated,
but the resulting equation is unusual; it cannot be expressed as
concentration as a function of time. However, it can be manipulated
to give t = F(concentration). It would be easy to fit data in pro
Fit to this equation. But...
But this is wrong statistically, isn't it? In my case, t is the
(experimental) independent variable, and the error on it is very
small compared to the error on the (experimental) dependent variable
- the concentration. Isn't it improper to use standard fitting
algorithms in such a case, i.e., isn't it implicitly assumed that the
error on the independent variable is negligible? Unfortunately, I
can't transform my function so that concentration = f(t).
Any thoughts on how to obtain valid fits with valid errors on the
parameters?
ciao,
Bruce
More information about the proFit-list
mailing list