| Author |
 Topic  |
|
|
ideal4ever
USA
2 Posts |
 Posted - 08/15/2014 : 05:09:33 AM
|
Origin Ver. 9.1.2
Operating System: Windows 8
I am doing Peak Integration and using Fitting (Pro) for fitting my baseline. Unfortunately it seems like in the "Peak Integration" module you can only choose the fitting function and assess the goodness of the fit by visual inspection. This is unlike Non linear curve fitting where you can assess the goodness of fit by statistical measures like adjusted R-square or chi-square. Although, in the non linear curve fitting there is no option for setting baseline "anchors" around your region of interest! So my question is that is there anyway to have a more rigorous baseline fitting (meaning being able to access goodness of fit by statistical measures e.g. Adjusted R-square etc.) when using Peak Integration? To put it another way is there a way to do a rigorous non linear curve fitting in the "baseline fitting" step of the "peak integration"?
|
|
|
greg
USA
1378 Posts |
Posted - 08/15/2014 : 4:06:01 PM
|
The fit is "rigorous", it's just that we never considered anyone would need the statistics of the baseline fit.
Here's how you can get them:
Go to your FitPeaks Report page and expand the Baseline Mode branch.
Right-click on "Baseline Anchor Points" (or click the drop-down button to the right) and choose "Create Transposed Copy as New Sheet".
Now use the nonlinear fitter and pick the Baseline function you used and set your data X and Y to the correct baseline X and baseline Y.
When completed, you should have the same baseline parameter values, but now you will have the statistics as well. |
 |
|
|
ideal4ever
USA
2 Posts |
Posted - 08/15/2014 : 4:48:42 PM
|
| Dear Greg, Thank you very much for taking time responding to this comment. Just a quick note regarding my need for statistics of baseline fit: I am dealing with a exponentially decreasing signal with orders of magnitude change in the Y value and low intensity peaks. The result of integration of the low-intensity peaks normally corresponds to an absolute value for a physical phenomenon, accuracy of which depends very much on how the baseline is removed (actually this is one of the factors affecting the accuracy), and that's why I need to know how "good" is the baseline subtracted. Now from what I understood you are suggesting to export the anchor points' coordinates (X, Y) and basically do a separate NL curve fit on them using the same fitting function I used in Peak INtegration and extract the fitting statistics. This is brilliant idea and I believe I could do other way around, i.e. first do NL curve fitting on some anchor points and use same anchor points and function in the Peak Integration module (but I already know the fitting statistics). Except one thing that I still haven't figured out: When doing NL curve fitting on some anchor points, I use "weighting" on the points around my area of interest (low intensity peaks) to make sure that fit is optimized (visually) - Now is there a way to carry this "weighting" on certain anchor points to "Peak integration" when Loading the anchor points from NL curve fit? |
|
|
|
greg
USA
1378 Posts |
Posted - 08/18/2014 : 09:29:33 AM
|
The Peak Analyzer does not have an option to include weight on the baseline part of the fit.
So it seems your best approach would be to use the baseline function to fit (with weighting) a set of baseline points. The Peak Analyzer can then use the resulting fit as the baseline.
If you do the fit from a graph which by default shows a table including Reduced Chi-Sqr, then you can use Data : Move Data Points to push your anchor points around and see if you can improve Reduced Chi-Sqr or alter Y or Y-error values in the anchor points worksheet. |
|
|
|
johnmering
New Zealand
1 Posts |
Posted - 11/05/2015 : 4:52:18 PM
|
Hi-
I'm working with laser spectroscopy data from an LGR isotope analyser. I'm trying to subtract a baseline and integrate peaks. Attached is an image from a run.
My goal is to implement a baseline correction at each peak. Note: there are 13 peaks. I want to select a control point on either side of each peak in the "user defined" section for baseline correction. This can be done automatically using the AutoFind mode for finding baseline anchor points. From there, I can manually adjust the points to get them to relatively the correct location.
Ideally, I'd like to set it so that each peak is only corrected to the baseline control points that are placed immediately before and after. Is this possible? Is there a straightforward way to consistently implement a baseline correction for each peak in runs like this?
Thanks so much!
 |
|
|
|
cdrozdowski111
USA
247 Posts |
Posted - 11/12/2015 : 10:05:56 AM
|
Hi johnmering,
You might consider using our Integrate Gadget on the graph. It allows you to create a resizable and draggable Region of Interest box around individual peaks and set a baseline only within that region.
On the top-right of the gadget is an arrow button. Click that button to open the Preferences dialog where you can make many settings including how the baseline is handled and what you want to output for the integration of each peak. In your case, you'd want to set it as "Curve within ROI".
Also in the Preferences dialog, you can tell the gadget how you want to create output. Once you are ready to output information for a selected peak, simply click on the arrow button and select "New Output".
Give it a try and see if it fits your needs.
~Chris
|
Edited by - cdrozdowski111 on 11/12/2015 10:07:03 AM |
|
|
| |
Topic |
|
goodness of baseline fitting in "peak integration"