I tried to fit an error function to data points. The erf has the form:
y = a - a * erf(x/b)
However, it looks like the built in nag_erf(x) term does not support a parameter inside the erf-term? Or am i mistaken? Maybe someone can give me an advice to solve that problem.
I tried this in 7.5SR5 and it worked fine for me. Note that you could use erf() or nag_erf(). The two are different in that the erf() function calls internal Origin function that has existed since before NAG was added. nag_erf() calls the NAG library. Both should yield same results.
So I created an FDF with the function y = a - a * erf(x/b); with Origin C turned on in NLSF. I could then simulate with the function to generate some data, and then fit that data with some changed parameters and I could retreive the values I used in the simulate dialog.
Well thank you very much for your quick answer. I really don't know what exactly went wrong - because now it works just fine as it should. Maybe it was just one of these long days at work, when things go wrong...