| T O P I C R E V I E W |
| zubson1 |
Posted - 11/18/2012 : 3:00:13 PM
Origin Ver. and Service Release 8.5.0 SR1
Operating System: Windows 7 Ultimate 64x
I'm using origin C code from tutorial "fitting with convolution"
I have dataset A(x) and B(y) from experiment, and I know that I need to use Gaussian blur on fitting function.
The fitting function is y=a*(x-E)^1.35, where a and E is parameters
I need make a convolution with Gaussian with width = 0.5
so g=b*exp(-(x-xc)^2/2w^2), where b is parameter, x is from dataset A, w is width, xc is SECOND independent parameter
so I need to do integral g from -inf to inf, and nex make convolution with y
I read this tutorial, and I think it is ok, but I must generate response (gauss) for only one xc, and I get very bad result of this program (I copy code from tutorial and change only the function y)
How to make fitting with convolution?
The fitting function in one piece is:
Y=integral(-inf,inf) [a*(x-E)^1.35 ** b*exp(-(x-xc)^2/2w^2)]dxc
** - is matemathic convolution two functions
so I have independent parameters: x, xc
dependent parameter: y
parameters: a,b,E
constans: w |
| 1 L A T E S T R E P L I E S (Newest First) |
| Penn |
Posted - 11/19/2012 : 01:55:07 AM
Hi,
I don't think your case is similar to the tutorial "fitting with convolution". In the tutorial, the response is real data, but in your case, you seems to use the expression as response. Also, there are other issues.
1. a*b should be considered as one parameter. If you insist using a and b two parameters, you need to fix one when fitting.
2. If xc is the second independent parameter, there should be one dataset for it, just like dataset A for x as you mentioned.
3. In your "one piece" fitting function expression, xc is the integration variable, after integration, xc will not exist in the function anymore, then xc cannot be an independent parameter, right?
So, if you can send the data to technical support (click the link in the top-right corner of the page) and provide more details, we can try to see whether there is a solution for you. When sending the email, please refer to this post.
Penn |
|
|
