| Author |
 Topic  |
|
|
Chao
Cambodia
Posts |
 Posted - 07/29/2005 : 11:41:53 AM
|
Origin Version (Select Help-->About Origin): 7
Hi,
I have several datasets of bins (60-1200, bin size 30) and frequency (%) for each bin. I am trying to fit an exponential decay function (seems to fit the data best by eye, I am tried all of the possible ones the fitting wizard offers) but I keep getting errors like "Error! Parameter(s) y0, A1, t1 is (are)not properly initialized. Check their values". I tried guessing some parameters but as soon as i try to iterate it kicks out one of the parameters again and gives the same error. Removing all empty bins (zeros) from the data set, allowed me to fit it in Excel but the fit is bad. The first (large) frequencies in the bins are statistically more accurate and have to weighted more in the fit. I am thinking maybe my data is just not good enough, but shouldn't Origin anyway be able to fit it?
Any help/advise is greatly appreciated.
Chao |
|
|
Chao
Cambodia
Posts |
Posted - 07/29/2005 : 5:19:32 PM
|
please help. I need to give a presentation on my data and need those fits to compare it properly.........
thanks so much |
 |
|
|
Mike Buess
USA
3037 Posts |
Posted - 07/29/2005 : 6:26:21 PM
|
If your bin size is 30 then your wks only has 40 or so rows. Why not copy/paste the wks data to another post? It's impossible to tell what's happening from your description alone.
...And tell us the name of the function you used. (There are several exponential decay fns.)
Mike Buess
Origin WebRing Member
Edited by - Mike Buess on 07/29/2005 6:39:54 PM |
|
|
|
Chao
Cambodia
Posts |
Posted - 07/30/2005 : 12:27:56 AM
|
60 10.67
90 11.35
120 8.05
150 7.81
180 6.79
210 5.30
240 5.02
270 5.33
300 4.17
330 4.61
360 2.89
390 3.66
420 3.39
450 2.73
480 2.27
510 0.35
540 0.73
570 0.77
600 2.45
630 1.29
660 1.80
690 0.94
720
750 0.51
780 0.53
810 1.11
840
870
900
930 0.64
960 1.98
990 1.36
1020
1050 0.72
1080
1110 0.76
1140
1170
1200
Where there is no data, the bin is empty (0). I removed the zeroes as this allowed me to fit it in Excel. The lower bins (say below 210) should be weighted higher in a fit as they are statistically more significant. I tried fitting with ExpDec1,2,3 as well as ExpDecay1,2,3. |
|
|
|
Mike Buess
USA
3037 Posts |
Posted - 07/30/2005 : 09:33:44 AM
|
Hi Chao,
I imported your data to Origin 7.0 SR4 and got no error messages from the Fitting Wizard using ExpDec1 with the wizard's defaults. However, when I tried to apply statistical weighting the wizard did not complain but reverted to no weighting nonetheless. (Even though I had deleted the cells with zero value as you prescribed.) In any case I'm much more comfortable using the Advanced Fitting Tool and suggest you do the same...
1. Starting with a plot of your data select Analysis > Advanced Fitting Tool.
2. Click the More... button if you see it.
3. Select Function > Select on the Fitter's menu and choose the fitting function. (I chose ExpDec1.)
4. Select Action > Dataset and select your dependent (Y) and independent (X) datasets.
5. Select Scripts > Parameter Initialization and check the 'Enable Automatic Parameter initialization' option.
6. Select Scripts > After Fit to choose how you want to output the results. (I generated fit curve and pasted parameters to plot.)
7. Select Options > Control and pick a weighting method. (I used Statistical with the 'Scale errors...' option unchecked.)
8. Select Action > Fit. Parameters should already be initialized. (I did not adjust.) Click the 100 Iter. button until convergence occurs.
9. Click Done to finish.
This is what I got...
Mike Buess
Origin WebRing Member
Edited by - Mike Buess on 07/30/2005 09:44:12 AM
Edited by - Mike Buess on 07/30/2005 10:01:02 AM |
|
|
|
Chao
Cambodia
Posts |
Posted - 08/01/2005 : 11:56:31 AM
|
I was still not able to do it (same error). I just noted though that you had SR4 while I didn't. So I just updated my Origin and now it works just as you describe. THANKS A BUNCH!
|
|
|
| |
Topic |
|
problems fitting a data set