| Author |
 Topic  |
|
|
tony_lincoln
USA
Posts |
 Posted - 08/18/2009 : 01:29:09 AM
|
Hi,
I have to use sigmoidal funciton to fit several curves, say, 3 curves. The sigmoidal funciton is: f(x) = 1/(1+exp(-k*x)). So after fitting, I will get k. After curve fitting, I just simply get the average of 3 k values. That is all.
But, in Origin, the sigmoidal fitting was finished by Boltzmann function, which is y = A2 + (A1-A2)/(1 + exp((x-x0)/dx)). So what is k, in this case?
I guess, 1/dx should be k. for example, if dx = 2.0, then k = 5. so I should use 1/dx of sigmoidal function as the k in sigmoidal. Then get the average of 3 k values.
Am I correct? sorry for this question, and your message is highly appreciated.
Thanks.
Tony |
|
|
larry_lan
China
Posts |
Posted - 08/18/2009 : 02:53:11 AM
|
Hi Tony:
The Slogistic1 function under Growth/Sigmoidal category looks more like the function you want.
In most cases, we add a horizontal offset for built-in functions, so you can see most of the function were defined like (x -x0), where x0 is the offset. If you dont want this parameter, you can Fix it. (If you use the boltzmann function, you should also fix A1, A2)
It's OK to calculate k from -1/dx, the value is same, but the Error is different (You can't compute the error by Kerr = -1/DXerr ). So I recommend you creating a user-defined fitting function for your case.
Thanks
Larry |
 |
|
| |
Topic |
|
|
|
sigmoid curve vs. Boltzmann funciton (curve fit)