The goal of Multifit is to provide a convenient way of calculating fits with shared parameters over multiple data sets.
The actual fitting is all done using the iminuit
package, and I recommend that you familiarise yourself with their basic tutorial.
Install from PyPI repository (recommended)
pip install multifit
To install the developement version, clone this repository, cd
to its root directory and run
pip install .
Import the UI
class and initialise it with a list containing the paths to the spectrum files.
import multifit.UI as fitUI
paths_to_spectra = ['path/to/spectrum1/spec1.asc', 'path/to/spectrum2/spec2.asc']
UI = fitUI(paths_to_spectra)
Then, either call UI()
to run the program with the default configuration, or call the methods in UI
manually to have full control over the fit. This will be discussed below.
If the get_input
method is called, the program will attempt to read certain arguments from the command line, and prompt the user to enter them if this fails. These are
interval
: the interval of the data to fit.
m_init
: approximate values for the peak positions.
savename
: the location to save the results, with no file extention.
Once the fit is complete, the results are contained in the Minuit object UI.m
. See the iminuit documentation for information on how to extract it.
Working examples are given in doc/examples/.
In the default configuration, multifit can fit any number of peaks using Gaussian functions on a linear background. The mean and standard deviation of each peak is constant across all spectra. The peak areas and background parameters are fitted independently for each spectrum.
Each step of the fit, as well as the variables involved, can be accessed through the UI
class. The following contains a description of some variables you may want to change. For more details, see doc/reference.md
clik to expand
-
self.function_string
: This is a string which, when executed, defines the function to be fitted to the data. The function should takex
as the first argument, followed by single parameters to fit. The program does not support array parameters, because the selection of common parameters is name based. The default string is created inmake_Gaussian(self)
. This method may be usefull as a template when creating your own fit functions. -
self.Chi2
: This is the ย ฯ2 cost function which is minimised byMigrad
. It is defined inmake_chi2(self, fit_function: callable
) as
self.Chi2 = self.Fit.Chi2(self.interval, fit_function, [*self.listall('A'),'slope','offset'])
You will only want to change the last parameter, which is a list of the parameters to be fitted independently for each spectrum. By default these are the areas of all peaks and the background parameters.
-
self.initial_values
: This is adict
containing the initial values for all parameters of the fit. For those parameters which are fitted independantly, that is one value for each spectrum. The default is defined inset_initial_values(self)
. It is recommended to copy and modify this function to generate custom initial values. -
self.m
attributes:self.m.limits
: the limits of the fitted parameters, by default the peak areas are positive.self.m.fixed
: can be used to fix the value of fitted parameters.False
by default.self.m.values
: the current values of the fitted parameters. See the iminuit documentation for more information.