| Author |
 Topic  |
|
|
branci
Brazil
4 Posts |
 Posted - 11/19/2002 : 9:02:08 PM
|
| I need of the function Y = P1/(1 + P2/x + (x/P3). This function must be a hyperbolic function. I try... and try... and try... plot this function in Origin, but I never get. Help me. I thing that there is an syntax error, but I am not sure. |
|
|
Mike Buess
USA
3037 Posts |
Posted - 11/19/2002 : 10:34:58 PM
|
Perhaps you just made a mistake posting the equation, but your parenthesis don't match up. Which of these is the correct equation?
1) y = P1/(1 + P2/x + x/P3)
- or -
2) y = P1/(1 + P2/x) + (x/P3)
Mike Buess
Origin WebRing Member |
 |
|
|
branci
Brazil
4 Posts |
Posted - 11/20/2002 : 3:04:45 PM
|
| I try these two forms but it doesn't work. My origin version is 6. If someone could help me please DO IT! |
|
|
|
branci
Brazil
4 Posts |
Posted - 11/20/2002 : 3:05:49 PM
|
| The correct form is one. |
|
|
|
easwar
USA
1964 Posts |
Posted - 11/20/2002 : 4:45:19 PM
|
Hi,
Maybe you can do the following and then report further problems here:
1> Plot your data
2> Start NLSF tool (menu item: Analysis->Nonlinear Curve Fit...
3> In NLSF, choose the menu item Function->New
4> Type in a name for your function
5> Leave Type drop-down as User-Defined
6> Change the "Number of Parameters" drop-down to 3
7> Leave "User Defined Param Names" box unchecked
8> Leave "Indep var" and "Dep var" edit boxes as x and y
9> In the function edit window, type
y = P1/(1 + P2/x + x/P3)
10> Change the "Form" drop-down list to Equations
11> Click "Save" button to save function
12> Go to NLSF Action->Fit menu
13> Click on "Active Dataset" button in the dialog box
14> Now assign initial guess values for P1, P2, P3
15> Click the 1Iter or 10Iter button
Do you now get a fit?
If not, do you get an error message?
If yes, what error message do you get?
Easwar
OriginLab. |
|
|
|
branci
Brazil
4 Posts |
Posted - 11/20/2002 : 5:24:27 PM
|
I make all you said, but my graph is a line and no a hyperbole. My data is
X Y
0,05 1,8
0,1 6,4
0,2 16,6
0,6 31,79
0,8 52,7
1 69,8
2 83,9
5 112,3
10 129,7
20 127,4
30 62,2
50 40,2
If you fit using a hyperbl function... I get a hyberbole graph. But if a use my function I don't get my graph. The hyperbole should follow the scatter.
|
|
|
|
Mike Buess
USA
3037 Posts |
Posted - 11/20/2002 : 7:21:09 PM
|
I fit your data and equation in Origin 6.0 and success depended a lot on the initial guesses. First I tried P1=P2=P3=1 and got a straight line with large negative Pn values. Then I tried P1=130 (close to maximum y-value) and P2=P3=1 and got a reasonable looking curve. After four or five clicks of the "10 Iter." button I ended up with these parameters...
P1=263,44546
P2=3,60752
P3=13,41176
Mike Buess
Origin WebRing Member |
|
|
|
greg
USA
1378 Posts |
Posted - 12/06/2002 : 09:47:13 AM
|
I guess we can't emphasize enough ...
USER DEFINED EQUATIONS REQUIRE GOOD INITIALIZATION
The (default) Levenberg-Marquardt algorithm is particularly susceptible to 'wandering' in it's parameter space. Sometimes the Simplex method is a bit better when initial guesses are bad, but this depends on the function.
|
|
|
| |
Topic |
|
|
|
Neqw Function....Help meeeeeeeeeee!