"bad svar" in Multipeak Fitting 2

Share icon

Somehow the Multipeak Fitting 2 package stopped working recently. I have attached an example experiment.

Specifically, after loading the package, choosing y & x waves, "continue", I see below a 'bad svar' message, and any subsequent analysis simply does not work. I am not sure what happened since I just used it a couple of days ago on the same file. Graph0_1.pxp (11.38 KB)
It looks like the experiment file you posted is the result of choosing File->Save Graph Copy. If it isn't the exact experiment file you are working on, then this answer may not be correct. However...

The Packages data folder in your experiment file completely lacks the data that Multipeak Fit 2 uses to keep track of the "set" number and a bunch of other things. The fact that it has a MultiPeakFit2 data folder makes the package believe that it should have a bunch of information that is missing.

It may be that your best course of action would be to move the data you want to fit into a completely new experiment file and start again.

You can recover your ability to use Multipeak Fit 2 by doing this:

1) Select Windows->Procedure Windows->Procedure Window and remove this line:
#include <Multi-peak fitting 2.0>

2) Bring up the Data Browser (Data->Data Browser) and open up the Packages folder. Remove the Multipeak Fit 2 data folder. If you want to keep your nicely formatted graph, you will need to move the Peak 0 and Peak 1 waves to a new location.

3) If the Start Multi-peak Fit panel is still open, close it. You can find it by choosing Windows->Other Windows->Start Multi-peak Fit.

Now you can re-load Multipeak Fit 2 and try again.

It just occurred to me that you may have gotten into this by choosing File->Save Graph Copy, thinking that it would allow you to make an experiment file with just that data and graph in order to continue working with Multipeak Fit 2 using that graph. In fact, the dependence on a private data folder with data that the package depends on is one of the big weaknesses of procedure packages.

John Weeks
WaveMetrics, Inc.
support@wavemetrics.com

June 24, 2015 at 10:21 am - Permalink

Thanks for the response. I did move the raw data to a new experiment. Now I don't have the bad svar error, but the fit just does not make sense. Or, in other words, there is no fit. I have saved this new experiment as attached.
test_2.pxp (48.03 KB) test_2.pxp (48.03 KB) test_2.pxp (48.03 KB)

June 24, 2015 at 11:17 am - Permalink

You are experiencing numerical instability. The fit works fine if you scale your X values up. I have attached a copy of your experiment in which I multiplied your X values by 1E6 before fitting. I fit Gaussian peaks (set 4) or ExpModGauss peaks (set 3). The ExpModGauss makes a slightly better fit. It may be that there are actually more than two peaks that are heavily overlapping. Only you can say what is appropriate to your data and the physics involved.

John Weeks
WaveMetrics, Inc.
support@wavemetrics.com
test_JW.pxp (159.41 KB)

June 25, 2015 at 10:21 am - Permalink