ExpModGauss fitting

Share icon

Thu, 02/12/2015 - 04:15 am

Hi,

I have a big assymetric Raman peak and i need to perfrom some fitting to define the position......Its a material produced in a specific way (amorphous calcium phosphate) so we are not entirely sure about the chemistry.....I can either assume some contribution of smaller peaks to the big one and perform gaussian fitting for all or perform some assymetric fitting to the big one only.....in any case the position hardly changes (maximum 0.5 wavenumber difference). When i perform assymetric fitting, it gives the location of the ExModGauss, but it also gives me the location of the actual Gaussian. However, i do not get the standard deviation for the amplitude and the position of the ExModGauss (it says not available)......Any ideas why is that? It is not so important since i get the location of the ExModGauss but i was just wondering.

Best regards,

Konstantinos
This is not answering your question, but I have found that in situations where the peak is asymmetric and I don't want to impose a parameterised function on its shape, the Barycentric mean (centre of gravity) can be a useful measure of the peak position - you are using all of the intensity of the peak to determine the peak position. It is less useful, however, if the peak is not well isolated from neighboring peaks. (Bmean = sum(x * intensity(x)) / sum(intensity(x), calculated over the range of the peak)
hth,
Kurt

February 12, 2015 at 06:20 am - Permalink

Hi Kurt,

Thanks very much for this useful information......I can just try it and see how well it correlates with the modified gaussian that i used.....As i mentioned, ACP has been studied a lot but my samples are produced in a specific way so we are not quite sure about the chemistry and this makes it more difficult to define if there is convolution or not.....the assymetry often indicates convolution, but i do not see any obvious shoulders so i also try assymetric fitting to the individual peak. One question......what software do you use to find the Barycentric mean? Thaks very much

Kind regards,

Konstantinos

February 12, 2015 at 06:40 am - Permalink

The fit coefficients for a given peak shape don't necessarily include all the information you want. Multipeak Fit 2 fills in "derived" information by computing it from the fit coefficients after the fit is done. If there is an analytic way to compute such derived quantities, then propagation of errors is used to compute the standard deviation from the errors of the raw quantities (some of those expressions get really complicated!).

I am not aware of any analytic way to get the location or amplitude of the real maximum of the ExpModGauss peak shape, so it is computed by numerical methods (Optimize for the peak location and amplitude, FindRoots for FWHM). Such a technique leaves me also unaware of any good way to get the standard deviation from the raw fit coefficients short of some sort of Monte Carlo technique. If you know more than I do, please tell me!

John Weeks
WaveMetrics, Inc.
support@wavemetrics.com

February 12, 2015 at 09:11 am - Permalink

Konstantinos Chatzipanagis wrote:
One question......what software do you use to find the Barycentric mean? Thaks very much

Hi Konstantinos,
I use Igor (of course!) - just write a simple for-loop to do the sums over the desired range.
Have a go - if you are struggling then I can try to help you out.
Regards,
Kurt

February 15, 2015 at 11:44 pm - Permalink