I used above objective function to fit my data in matlab, but the results are very sensitive to initial values,
could anybody please give me some advices on what should be noticed when doing the fittings in mathcad?
in the equation, Y~X are obtained from experiments, Sp and rsp are parameters to be estimated
Please attached the Mathcad worksheet.
Mike
Hi, Mike
For some reasons, I cannot upload the raw data this time, but I can use a similar problem as an example.
see below
Data from experiments are plotted by (X,Y)
and I wanna use a normal distribution function as an objective function, but when I do this in mathcad, I often got complex σ and μ , and the results are initial value sensitive. see following process
attached are the mathcad 15 worksheet and its pdf counterpart
but when I apply this method to another problem, there are no such strage results, see below:
It works if you omit the first data pair with the x=-1. Is that pair real?
Yes, I noticed this last night, everything worked well when I used Re() to get the final results.
but I dont know why.
When I was a student, I encountered the similar problem in matlab: when I used a set of ODE to fit some data for obtaining concerned parameters. and the solution is also the same