Data fitting in LSE method 1

[iqbal67]

Hello

I am attaching the details of the problem here. Please help me to solve the problem with LSE fitting. I have 3 fitting parameters and i mentioned the measured data with which i want to fit my calculated data.

I would really appreciate your prompt response. If more information needed, i would immediately respond.

 

[Occupant]

I think, you don't have enough data points to do anything decent with it.

 

[IRstuff]

I don't get it. If your log-lin plot implies a straight exponential function, why not just use line(X,Z) where Z=ln(Y)?

http://files.engineering.com/getfil...41d-9b3c6876a541&file=LSE_fitting_issue1.xmcd

TTFN

FAQ731-376
Chinese prisoner wins Nobel Peace Prize

 

[IRstuff]

The LSE I get with that fit to Y is 2.742 x 10^-33, assuming that's what you're looking for.

TTFN

FAQ731-376
Chinese prisoner wins Nobel Peace Prize

 

[iqbal67]

Hi

I understand that an exponential fit is fine. However, my primary intention is to find out the parameters that i mentioned. I must vary them to get the parameters as reasonable as possible. These extracted fitting parameters will be used to obtain another modeling equation. Please suggest me any good way of varying these parameters for LSE fitting. The data might not fit exactly, but i want to know how can i vary these parameters so that Mathcad will do the iteration for me to obtain minimum error. In MS Excel i can do it, but it does not converge well.

Thanks

 

[Occupant]

So, why don't you use regress as I suggested. Just change the degree of polynomial fit to 2, you get a better fit.

 

[IRstuff]

You can probably use Minerr or Find with the equation expressed as a vector equation.

TTFN

FAQ731-376
Chinese prisoner wins Nobel Peace Prize

 

[Occupant]

The problem seems to be that varying the three coefficients doesn't lead to an improvement in the solution - other than screwing it up totally if driven into extremes.

 

[iqbal67]

I have found a way of fitting the data in Excel by using the solver function and using GRG nonlinear method (Office 2010). It gives some sort of result, by varying the 3 fitting parameters and doing iterations for a long time. I am looking for something similar to that, as i don't know Excel's algorithm!

 

[Occupant]

Maybe just a stupid question, but you have defined all these vectore and then you don't use some of them like Rch and, by extension, Vch and Vg

 

[iqbal67]

@Occupant:
Those data are part of my next calculations, i did not remove those. This is a very big modeling calculation, i have just shown the initial step. After i get this correct, i can proceed to next steps. I understand that these are confusing, but i mentioned clearly what i want to vary and what is my goal, please look into that again. I would say this is a kind of optimization problem rather than exact data fitting. I need to optimize the fitting variables.Thanks.

 

[Occupant]

I am actually better than you, unless you've made pogress in the meantime - so there. I know, it still needs some tweaking, ran out of time.