| T O P I C R E V I E W |
| helga55 |
Posted - 11/16/2006 : 2:48:00 PM
Origin Version 7 SR4
Operating System:Win 2000
Please bear with me but what's the difference between NLSF Wizard (Analysis->Non Linear Curve Fit ->Fitting Wizard) and what seems like the simpler Peak fitting Module?
Thanks
Helga |
| 19 L A T E S T R E P L I E S (Newest First) |
| larry_lan |
Posted - 12/05/2006 : 08:41:05 AM
Hi:
Just like Zach mention above, PFM just do non-linear curve fit for replica peaks. So we need to give initial value for the peaks. The initial values may affect the result if it is unreasonable. And the Peak Edit Control step will change the initial value.
Larry
OriginLab Technical Services |
| helga55 |
Posted - 12/04/2006 : 2:10:19 PM
I am getting there but its a slow process!
Here is another Q: If I fine tune the peaks with the Peak Edit View Box (PFM) and bring the red line (theoretical) closer to the black line (actual data) then my final fit should be better shouldnt it? Also if I dont do this and let Origin do the FIT would my parameter values be different from what they could be if I had taken the trouble to fine tune the line before I hit FIT? I am asking the latter because I feel that PEAK EDIT VIEW BOX is just an initial estimate.
Thanks a lot
Helga |
| larry_lan |
Posted - 11/29/2006 : 8:13:06 PM
Yes, not only A but also w can be negative, and you can always set constraints when you found unusual values.
Larry
OriginLab Technical Services |
| helga55 |
Posted - 11/29/2006 : 09:29:33 AM
Thanks a lot Larry. so the carry home message is that the nature of the data may give a negative value for A and in order to prevent that one sets the lower bounds to zero. So just glancing at the original data may not tell you in advance if this will be so - I would have to run the PFM, then find negative values if any and then go back to the Parameter file and set the bounds. Is that right?
Thanks once again
Helga |
| larry_lan |
Posted - 11/22/2006 : 9:57:27 PM
Hi Helga:
The manual just want to show that you can add linear constraint in PFM. We used to set bounds when there overlay peaks. For example, the following red peak is composed by the two green peaks.
We know that the area of a peak should not be negative, however, if the parameter initial value is bad, we may get a negative area. Look at the Gaussian fitting function:
If A is nagetive, we can get a nagative peak like this:
This result is correct mathematically, but wrong in fact (nagetive area). So adding a linear constraint will optimize the fitting procedure. If your peaks looks OK, you can just ignore this steps.
Larry
OriginLab Technical Services |
| helga55 |
Posted - 11/22/2006 : 4:50:59 PM
Maybe my curve fitting statistics is not up to date so forgive me (!) but I am trying to run the example in my Peak Fitting Module and I couldnt understand one thing. The example is the PFM_Examples.opj and I hope its still in use as this manual came with ver 7. When I run the example and reach the Peak Edit Control Wizard section I am asked to set the lower bound of Peak #2 to zero 'to prevent the area from going negative during the iterative procedure' I couldnt understand why this would happen as the peak looked okay to me - however I would like to know why it could be so as I am getting negative values with my own data.
Thanks a lot
Helga |
| helga55 |
Posted - 11/22/2006 : 10:33:17 AM
I think you are very right Zachary. I thought the inital parameters were the fitted ones although in retrospect that was not smart on my part! Thanks a lot for the information.
Helga |
| zachary_origin |
Posted - 11/22/2006 : 10:28:13 AM
I mean the Center and FWHM of the peaks in parameter display sheet are only initial values for the parameters of peaks. After fitting, these will be changed according to the algorithm of PFM. So I cannot see any necessity of knowing the exact value of the inited parameters. And I am doubting you have mistaken the initial parameters as fitted parameters. Right? maybe things are not as I imagined.
Zachary
OriginLab Technical Services. |
| helga55 |
Posted - 11/22/2006 : 09:05:07 AM
Thank you Larry and Zachary - the Click and then Right Arrow trick did it. A small question for Zachary - when you say "The fitter can get the right position for peaks after fitting (Click Fit button in Fit page)even there is a little deviation. So you do not need position the peak very exactly" do you mean I should finish the fit in my PFM then go to my Parameter display where I can change the Center and FWHM of each peak and then run the whole PFM again? That's the only way I have figured out how to change my parameters.
Thanks a lot for all your help.
Helga |
| larry_lan |
Posted - 11/21/2006 : 10:23:41 PM
A small trick. When you step to Define Peaks, you don't need to add any more peaks if you don't want to. Just click the peak (the magenta lines) and press the Right Arrow key on your keyboard once. Then the current marker position will appear. Don't worry to do this because pressing the arrow key will not move the marker position.
Larry
OriginLab Technical Services
Edited by - larry_lan on 11/21/2006 10:25:14 PM |
| zachary_origin |
Posted - 11/21/2006 : 9:22:10 PM
I am not sure if I understood your problem right. You meant the picked peak by the threshold may be not exactly what the position you needed? If you need ajust the peak location, you can go to Define Peaks page and click the peak and drag it right or left. Furthmore, these peak center locations are just for giving an init value for the fitter. The fitter can get the right position for peaks after fitting (Click Fit button in Fit page)even there is a little deviation. So you do not need position the peak very exactly. Maybe I had understood something wrong?
Zachary
OriginLab Technical Services. |
| helga55 |
Posted - 11/21/2006 : 3:32:00 PM
Thanks a lot Mike and Zachary - I think I will start with the easiest ie the PFM which is part of Origin 7.0
A small question - when i run through the steps of the PFM and come to Pick Peaks, i get several peaks depending on my chosen threshold. That's fine but how do I know at what wavebands they are? If I go to Add/Delete Peaks and place my marker on the peak, I do get an active window that shows me the X-Y coordinates but then I have ended up either adding another peak maybe slightly away from the one I am intersted in (if I did not place my marker in the exact same spot) or just deleted it. Right now I am looking at the Peak Parameter display sheet to see at what wavelengths these peaks are and then deleting them if I am not intersted in them. Is there a better way?
Thanks a lot for everything. This is a great forum.
helga |
| zachary_origin |
Posted - 11/21/2006 : 01:48:39 AM
quote:
Zachary, I'm pretty sure PFM also came with OriginPro 7.0. In fact, I still have a floppy disk entitled "Peak Fitting Module 7.0".
Sorry for that...
Mike Buess, you are right.
Zachary
OriginLab Technical Services. |
| Mike Buess |
Posted - 11/20/2006 : 10:00:11 PM
Zachary, I'm pretty sure PFM also came with OriginPro 7.0. In fact, I still have a floppy disk entitled "Peak Fitting Module 7.0".
Helga, If you are fitting several spectra that have one or more peaks in approximately the same positions you can also try my MultiFit addon on the File Exchange. Instructions are in the MultiFit project that comes with the download. If your spectra have multiple peaks see the Fitting with Replicas folder in particular. If you have any questions you can email me at <mlb@nmrtools.com>.
Mike Buess
Origin WebRing Member
Edited by - Mike Buess on 11/20/2006 10:18:21 PM |
| zachary_origin |
Posted - 11/20/2006 : 9:44:48 PM
As I posted the reply above, I found another thing. You mentioned you are using Origin V7SR4, then are you really talking about Peak Fitting Module? As I know, there is no Peak Fitting Module in Origin v7.
Zachary
OriginLab Technical Services. |
| zachary_origin |
Posted - 11/20/2006 : 9:21:43 PM
Seems you need do the fitting for several datasets?
As you can see now if you want to fit curves with multiple peaks, there are three ways:
NLSF Wizard (Analysis->Non Linear Curve Fit ->Fitting Wizard), Analysis > Fit Multiple Peaks -> Gaussian and Peak Fitting Module. For short, I will call them as NLSF, Fit Multiple Peaks and PFM respectively.
Now if you do not want to manually locate the peaks center, you have two choices (with Fitting Multiple Peaks, you can not do it):
1> programming in NLSF. As Mike mentioned, NLSF is more programmable. If lots of curves need be fitted, you can write some LabTalk scripts to perform this. You can search with "NLSF" in Origin Programming help (Help > Programming) or view some previous topics in LabTalk forum and Origin C forum. Try to write the scripts and then you will find these are very convenient. Any problem during your coding, just post them in the forum and you can get a quick reply.
2> Saving and Reading them in a file in PFM. In PFM, if the curves are similar, you can first manully fit with one and save the peaks info (in Results page of PFM) in a file and then read them (in Peak Finding page of PFM) for the other curves.
Zachary
OriginLab Technical Services.
Edited by - zachary_origin on 11/20/2006 9:41:05 PM |
| helga55 |
Posted - 11/20/2006 : 2:03:41 PM
Hi Mike,
I am trying to deconvolute the peaks in a spectrum of phytoplanton cells - so its mainly chlorophyll and other pigments.
From these two
http://www.originlab.com/forum/topic.asp?TOPIC_ID=4857
http://www.originlab.com/forum/topic.asp?TOPIC_ID=1107
discussion topics I now realize I should be plotting the absorption curve ie Plot->Line and then go to Analysis > Fit Multiple Peaks -> Gaussian whih prompts me for the Number of Peaks and Initial Half Width Estimate. Then it seems that I have to manually located the peaks that i want. Is there some way I can import them as a text file?
also can this same exercise be done with my Peak Fitting Module?
I am in uncharted waters so you will have to bear with me for a while!
Thanks a lot
Helga
|
| Mike Buess |
Posted - 11/16/2006 : 10:51:43 PM
Hi Helga,
PFM is better for multiple peaks... fits over 100 peaks with different functions. NLSF is more programmable.
Mike Buess
Origin WebRing Member |
| Deanna |
Posted - 11/16/2006 : 10:11:32 PM
Actually, this two tools are designed for different applications. PFM is more intuitive and easier to use. However, NLSF is more powerful. You can choose between them according to your need.
Deanna
OriginLab Technical Services |
