non linear curve fitting

[Maryam hosseini]

That is your VI that you gave me yesterday, It does not do the fitting properly.

[Maryam hosseini]

I am so sorry; I got a little bit confused with my data files as I have like 100 data files.

Could you ignore the last three messages, and just have a look at this VI that you gave me yesterday, It is working for that set of data, but why doesn’t it work for this new data set?

 

[Maryam hosseini]

Hi

It is working now. I am so happy now. Thanks alot.

You were very very helpful.

All the best

[altenbach]

I guess you found out to delete the last line so you don't get an extra zero. 😉 Here are a few more things you should consider:

• I always recommend to also graph the model using the parameter estimates. This way you can see if you're even close.
• Many times it is possible to calculate the initial estimates automatically since. In this case you could find the position of the max absolute y value and use its y value and x-position for a1 and a2, respectively. (this works for positive and negative a1 since your model does not have a significant offset.).

See modification "MOD2"

You have a signifincant systematic error, possibly caused by the fact that your data is actually not really a Lorentzian. An alternative explanation would be that your model is insufficient. For example if you would add a fourth parameter (a4) allowing for a small offset, the residue improves by about a factor of 10 (!).

See modification "MOD2-4params" (image below).

Message Edited by altenbach on 03-01-2008 11:31 AM

[Maryam hosseini]

Could you have a look at the part of the program that I have highlighted and explain it?

I can not really understand what you are trying to do in this bit of code. 

[altenbach]

1. This is a static reference to a VI used inside the fitting algorithm that calculates your function based on the x array and the parameters. It is called by reference at (4).
2. Here we look at the absolute Y data to see where the aboslute maximim is. We use the index to find the y and x value at that location.This will give us an estimate of the amplitude (A1) and position (A2).
3. We assume that the width paramtere (A3) is a constant at 1e6. We then form an array of these three values to automatically generate reasonable initial parameters to be used by the fitting tool.
4. Here we call the function by reference node to calculate how Y would look like using the parameter estimates. This way we can ensure that they are reasonable.

All clear?

Message Edited by altenbach on 03-02-2008 09:58 AM