How is Gaussian fitting done?

Question

Please suggest a website where I could get information about how to perform gaussian fitting on a set of data points.

Answer 1

try these

http://bima.astro.umd.edu/wip/manual/node11.html

http://www.ucl.ac.uk/~ucaptss/work/technical/gaussfit/index.html



I don't know if they are too complicated for you.

I think a guas fitting uses the sum of least squares.

what is your level of expertise

or at wikipedia

http://en.wikipedia.org/wiki/Least-Squares_Fitting - check out newton-gauss method

Answer 2

Gaussian Fitting







This method tries to fit one or more Gaussians to the data. In addition to the style (which must be ``gaussian''), the number of Gaussians to fit must be supplied. Additional parameters may also be supplied to help the fitting routine. For each Gaussian fit, there are three parameters used to describe it: the amplitude, the X-peak position, and the full width at half the peak amplitude (FWHM). Or as an equation:



where is the amplitude of the ith Gaussian; , the peak position; and , the FWHM. The resulting curve will be the sum of the N

This is waaaaaaay above my head, but I hope this helps?

Gaussians:



Note that if any of the supplied Gaussian terms are negative, they will be held fixed during the fitting.



When a fit is found, this method returns three coefficients for each Gaussian fit and the chi-square value of the fit.



6.2 Limiting and Displaying Fits

Answer 3

on your case you want to slot a gaussian with 4 parameters. those are DC offset (the heritage fee), area of height, height of height above the heritage, and classic deviation. installation a gaussian isn't so trivial. you possibly can take the log of the records and fit a quadratic it truly is an similar variety as installation the gaussian and that would want to artwork on your records, yet this has a pair of issues: One is that the incorrect mistakes is being minimized, because you reduce the sq. of the adaptation of logs to that end, and the different is that you may want to ought to eyeball the dc offset and then take care of what takes position even as pattern values are on the point of 0 above this dc offset, which could supply adverse infinity for the log values. you may want to easily subtract out the dc offset and then purely use relative height values that were > say 0.a million. you may want to then be installation utilising purely records that changed into heavily above the baseline. i'm assuming that you may workout a thanks to slot a quadratic. the purpose mistakes function you honestly want is E(offset,earnings,recommend,sigma) = y_i - (offset+earnings*exp(-(x_i - recommend)^2 / (2*sigma))), the position x_i and y_i is your records. the purely thanks to proper fit the gaussian is by using utilising non-linear least squares. this may be performed in matlab with the lsnonlin function, even with the very incontrovertible certainty that I keep in mind that you do not have get entry to to matlab. in actuality you should regulate the parameters {offset, earnings, recommend and sigma} to do gradient descent on your sum of squared mistakes Sum{E()^2} over the pattern set {x_i, y_i}. there is not any magic answer. Gauss-newton generation is possibly sufficient for this difficulty.