| Author |
 Topic  |
|
|
winflowers
USA
34 Posts |
 Posted - 02/20/2006 : 11:18:11 AM
|
Origin Version (Select Help-->About Origin): 7.5 SR5
Operating System: windows XP
Hi all,
I modified some codes copied from this forum to perform exponoential decay fit. Everything was going well until I tried to extrapolate some data. Here is an example, I would like to use the last 20 data in column b to do exponential decay fit and extend this fitting curve to some certain Y value.
Here is the code I am using:
void test(string wksName)
{
Worksheet wks(wksName);
string strDVar = wksName + "_b";
Dataset dsTime(wks,0), dsLum(wks,1);
double dHalfLum = dsLum[0]/2;
int iUpperbound = dsTime.GetUpperBound();
double dHalfLife;
using NLSF = LabTalk.NLSF;
NLSF.Init();
NLSF.Func$ = "ExpDec1";
NLSF.P1 = 0;
NLSF.P2 = 1000;
NLSF.P3 = 1000;
NLSF.V1 = 0;
NLSF.dataBegin = iUpperbound-19;
NLSF.dataEnd = iUpperbound+1;
NLSF.xBegin = dsTime[iUpperbound-19]; // set up x values
NLSF.xEnd = ??;
NLSF.xPoints = 50;
NLSF.FitData$ = strDVar;
NLSF.ChiSqrErr = 0;
NLSF.IterateEx(100);
NLSF.End();
double dDVal = dHalfLum, dIVal;
int iErr = nlfXfromY( dDVal, &dIVal, 1);
printf("The value of x where y = %g is %g", dDVal, dIVal );
}
Is it possible to set the xEnd value where certain Y is included? Or I need to perform mathematical function first to get the X value according to the parameters found from the fitting?
Any kind help is appreciated! |
|
|
Mike Buess
USA
3037 Posts |
Posted - 02/20/2006 : 12:52:49 PM
|
quote:
Or I need to perform mathematical function first to get the X value according to the parameters found from the fitting?
That's what nlfXfromY does. xBegin and xEnd are used with the makeCurve() method which you haven't used yet. (Perhaps it is called automatically by iterateX() or End()). Try something like this but I'm sure it will need some tweaking...
void test(string wksName)
{
Worksheet wks(wksName);
string strDVar = wksName + "_b";
Dataset dsTime(wks,0), dsLum(wks,1);
double dHalfLum = dsLum[0]/2;
int iUpperbound = dsTime.GetUpperBound();
double dHalfLife;
using NLSF = LabTalk.NLSF;
NLSF.Init();
NLSF.Func$ = "ExpDec1";
NLSF.FitData$ = strDVar;
NLSF.P1 = 0;
NLSF.P2 = 1000;
NLSF.P3 = 1000;
NLSF.V1 = 0;
NLSF.ChiSqrErr = 0;
NLSF.dataBegin = iUpperbound-19;
NLSF.dataEnd = iUpperbound+1;
NLSF.Iterate(100);
NLSF.Iterate(100);
NLSF.Iterate(100);
double dYVal = dHalfLum, dXVal;
nlfXfromY(dYVal, &dXVal, 1);
printf("The value of x where y = %g is %g\n", dYVal, dXVal );
NLSF.xMode = 5; // 1 and 5 work the same. Actually, you don't need this line at all!
NLSF.xBegin = dsTime[iUpperbound-19]; // set up x values
NLSF.xEnd = dXVal;
NLSF.xPoints = 50;
//NLSF.makeCurve("func"); // this is called by NLSF.End(). Use one or the other.
NLSF.End();
}
...Corrections marked in red work with my test data. The final message is this: everything you needed was in your original code but a bit of reordering was required to make it work.
Mike Buess
Origin WebRing Member
Edited by - Mike Buess on 02/20/2006 1:55:07 PM
Edited by - Mike Buess on 02/20/2006 2:48:02 PM
Edited by - Mike Buess on 02/20/2006 3:02:43 PM |
 |
|
|
winflowers
USA
34 Posts |
Posted - 02/21/2006 : 08:15:07 AM
|
Thank Mike for the codes!
Actually I have tested some similar codes but got the same problem. It seems that nlfXfromY(dYVal, &dXVal, 1) can only get the correct value within the fitting range, ohterwise you would get some unexpected result. In my case, I got 0 which is definitely wrong. That's why I raised this question.
What I am doing is just as I said in the first post to calculate the desired X value from Y value according to the parameters found from the fitting. Or is there any other solution? |
|
|
|
Mike Buess
USA
3037 Posts |
Posted - 02/21/2006 : 11:39:39 PM
|
Apparently you're right about nlfXfromY but nlfFit works outside the fitting range. You can try something like this to find dXVal at dY=dHalfLum without using the fit parameters directly. You could even use it to generate the fit curve itself.
double dYVal,dXVal = dsTime[iUpperbound-19];
double dXInc = dsTime[iUpperbound-18] - dXVal;
do
{
dYVal = nlfFit(dXVal);
dXVal += dXInc;
} while (dYVal>=dHalfLum);
printf("The value of x where y = %g is %g\n", dHalfLum, dXVal );
Mike Buess
Origin WebRing Member |
|
|
| |
Topic |
|
|
|
NLSF exponential decay