epsilonwaves in Fitting sums of Fit

Share icon

Hi, all,

I'm trying to fit the data with multiple peaks using 'FuncFit' as follows.

String myFunctions="{PeakLor, PeakTerm1, EPSW=PeakErr1}"
myFunctions+="{PeakLor, PeakTerm2, EPSW=PeakErr2}"
myFunctions+="{PeakLor, PeakTerm3, EPSW=PeakErr3}"
FuncFit {string = myFunctions} PeakSumData/D

Now I need epsilonwaves PeakErr(1-3) as standard deviations of variables PeakTerm(1-3) after fitting. However, somehow epsilonwaves do not change from the given values during fitting even though error values are changing in history. Is the keyword 'EPSW' not the right function? If so, how can I obtain epsilonwaves such as flag parameters '/E'? Thank you for any help.
Epsilon waves are not error waves. To help you more, we need a good description of what you are trying to do, beyond just "fitting data with multiple peaks". It is clear that you have misunderstood what epsion waves do, but it is not clear what you intended. What is it that you are trying to accomplish with EPSW?

Possibly you were trying to use weighting. The Sums of Fit Functions feature does not have a way to add a per-function weighting wave, as weighting goes with the data, not the fitting function. Use the /W flag as usual.

Finally, if you are fitting data with multiple peaks, is there a reason you aren't using the Multipeak Fit package we provide?

John Weeks
WaveMetrics, Inc.
support@wavemetrics.com

April 2, 2015 at 09:35 am - Permalink

John, Thank you for your answers.
Simply I want to get the coefficient errors after fitting the data with multiple peaks. For example, history box gives the coefficient errors (standard deviation) like

For function 1 PeakLor
Coefficient values ± one standard deviation
w_0 =124.22 ± 0.00232
w_1 =-28.658 ± 0.00125
w_2 =-0.26841 ± 0.208
For function 2 PeakLor:
Coefficient values ± one standard deviation
w_0 =128.73 ± 0.00428
w_1 =14.431 ± 0.00208
w_2 =0.3691 ± 0.124
For function 3 PeakLor:
Coefficient values ± one standard deviation
w_0 =183.35 ± 0.00116
w_1 =-42.509 ± 0.000566
w_2 =0.16125 ± 0.218

I cannot use Multipeak Fit package since the values of coefficient errors are used in my own program with a huge data set. I know how to get them in a single function by using flag '/E' such as

FuncFit myFunc, PeakTerm PeakData /E=PeakErr

but I don't know how to do in the case of "Fitting sums of Fit". Is there any way to extract those error coefficients shown in the history box?

Thank you,

IL.

April 3, 2015 at 08:10 am - Permalink

IL19 wrote:

FuncFit myFunc, PeakTerm PeakData /E=PeakErr

but I don't know how to do in the case of "Fitting sums of Fit". Is there any way to extract those error coefficients shown in the history box?

As I explained previously, the /E flag has nothing to do with error estimates. Even with an ordinary fit, the errors are reported via the wave W_Sigma.

In fact, that is how the error estimates are reported for a sums-of-fit-functions fit as well. I admit that it's not very convenient, but the error estimates for all the functions are all in the one wave, in the order in which the fit functions are listed in your command.

When I get time I probably should expand the documentation for sums of fit functions.

John Weeks
WaveMetrics, Inc.
support@wavemetrics.com

April 3, 2015 at 10:11 am - Permalink