Fitting data

[ptc-4749353]

Hi. I need help for fitting data. I am looking a function which can fit a kind of tailed function. see attached the data that I want to fit.

[ValeryOchkov]

May be the attach help you!

[ViktorKorobov]

May be so.

Viktor

[RichardJ]

What's the purpose behind fitting the data? Do you just want interpolated values? Do you want to understand something about the physics, chemistry, whatever, regarding this data?

[ptc-4749353]

ok I will explain.

when we are going to the left to the right, I want to have the area of second peak which is the signature of one chemical compound. I have many voltammogrammes were this peak increase with the concentration of the analyte. so I want to fit the data with a kind of two gaussian function. when I use a sum of two error functions the fitting is not good.

I hope that now you understand my problem

[ptc-4749353]

ok I will explain.

when we are going to the left to the right, I want to have the area of second peak which is the signature of one chemical compound. I have many voltammogrammes were this peak increase with the concentration of the analyte. so I want to fit the data with a kind of two gaussian function. when I use a sum of two error functions the fitting is not good.

I hope that now you understand my problem.

[ptc-4749353]

also, the function should have an electrochemical meaning. thank you

[RichardJ]

also, the function should have an electrochemical meaning.

That's why I asked. I'm not an electrochimist thoiugh, so you will have to provide us with a function that describes what is happening. Given the expected form of the function, we can show you how to fit it.

[ptc-4749353]

we want to used two gaussian functions.

[RichardJ]

we want to used two gaussian functions.

Maybe, but two Gaussians will not fit that data. Three Gaussians is better, but still not very good. Four gives quite a good fit.

[ptc-4749353]

Thank you.

But I don't understand why you don't use gaussian function as the one that I have attached and also please could you do this fitting using genfit function. I am new with mathcad, just three weeks that I have started to use it. I hope you undertans my position.

[RichardJ]

But I don't understand why you don't use gaussian function as the one that I have attached

Because that form of the function may be good for statistics, but it's not very useful for this case. First, it's normalized to unit area (i.e. total probability is 1). Notice that there are only two parameters: you can't vary the height independently of the width. Also, the width is defined as the standard deviation, which may be what you want for statistics, but outside of statistics it's not very useful. In the function I used the width is the full width at half height.

please could you do this fitting using genfit function.

I could, but I'm not going to. It's better to do least squares fitting using minerr, in the way I showed. It's easier to set up minerr, it's more versatile than genfit, and gives identical results.

[ptc-4749353]

OK . Thank you. now I try to learn fitting by leat squares. please could you show me some tutorials where I can started with linear function (the basic) and enhance difficulty. because this one is really tough for me.

[RichardJ]

If you have it, go to "help", "e-books", "data anlaysis extension pack", and look at the chapter on regression.