davebo's Journal: writing

I appear to be capable of 2 hours of actual work a day.

Read "HP & the Order of the Phoenix" last night. Worst. Harry Potter. Ever. Well, a little better than Chamber of Secrets - but not much better. 850 pages to learn Harry has to kill Voldemort, or Voldemort will kill Harry. Wow. Never saw that coming.

So I'm using these 2 programs (GREG & DDASAC) to fit some experimental data I have to a very simple model. These two programs have been worked on by a professor (the "S" from "BSL" for any chemical engineering geeks out there) and a number of his students for roughly 15 years . You'd think they'd work well by this point.

My model has 2 parameters, and is basically a sigmoid curve. The program is supposed to give me the values of these 2 parameters as well as a 95% confidence interval (using some statistical methods, I imagine) in which these parameters could fall. All I have to do is feed it my data, and an initial guess for the values of the parameters.

If I enter an initial guess of "20,000" for one of the parameters, and feed it my data, the solver tells me the "best possible fit" of that parameter is 28,000, plus or minues 9,000 (to give me the 95% confidence band). If I enter an initial guess of "40,000" for the parameter, and give it the exact same data, it tells me the best possible fit is now 46,000, plus or minus 40 (again, for that 95% confidence band).

I'm 95% confident that the numbers I'm getting out of this program are bullshit.