Interactive Curve Fitting – GUI Tools

This module contains the interactive GUI curve-fitting tools. They are based on Traits and TraitsGUI. Plotting is provided through the Chaco 2D plotting library , and, optionally, Mayavi for 3D plotting. The available models are those registered by the pymodelmit.core.register_model() mechanism.

FitGui – Interactive 1D model-fitting

Reference

class pymodelfit.fitgui.FitGui(xdata=None, ydata=None, weights=None, model=None, include_models=None, exclude_models=None, fittype=None, **traits)

Bases: traits.has_traits.HasTraits

This class represents the fitgui application state.

Parameters:

• xdata (array-like) – the first dimension of the data to be fit
• ydata (array-like) – the second dimension of the data to be fit
• weights –

  The weights to apply to the data. Statistically interpreted as inverse errors (not inverse variance). May be any of the following forms:

  • None for equal weights
  • an array of points that must match ydata
  • a 2-sequence of arrays (xierr,yierr) such that xierr matches the xdata and yierr matches ydata
  • a function called as f(params) that returns an array of weights that match one of the above two conditions
• model (None, string, or pymodelfit.core.FunctionModel1D instance.) – the initial model to use to fit this data
• include_models – With exclude_models, specifies which models should be available in the “new model” dialog (see models.list_models for syntax).
• exclude_models – With include_models, specifies which models should be available in the “new model” dialog (see models.list_models for syntax).
• fittype (string) – The fitting technique for the initial fit (see pymodelfit.core.FunctionModel).

kwargs are passed in as any additional traits to apply to the application.

dataChanged()

Updates the application state if the fit data are altered - the GUI will know if you give it a new data array, but not if the data is changed in-place.

getModelInitStr()

Generates a python code string that can be used to generate a model with parameters matching the model in this FitGui.

Returns:initializer string

getModelObject()

Gets the underlying object representing the model for this fit.

Returns:The pymodelfit.core.FunctionModel1D object.

weightsChanged()

Updates the application state if the weights/error bars for this model are changed - the GUI will automatically do this if you give it a new set of weights array, but not if they are changed in-place.

pymodelfit.fitgui.fit_data(*args, **kwargs)

Fit a 2d data set using the FitGui interface. A GUI application instance must already exist (e.g. interactive mode of ipython). This function is modal and will block until the user hits “OK” or “Cancel” - if non-blocking behavior is desired, create a FitGui object and call FitGui.edit_traits().

The following forms for input arguments are accepted:

• fit_data(xdata,ydata)

• fit_data(xdata,ydata,model)

• fit_data(model)

  This form requires a FunctionModel1D object that includes data

Parameters:

• xdata (array-like) – the first dimension of the data to be fit
• ydata (array-like) – the second dimension of the data to be fit
• model (None, string, or pymodelfit.core.FunctionModel1D instance) – the initial model to use to fit this data

kwargs are passed into the fitgui initializer

Returns:The model or None if fitting is cancelled or no model is assigned in the GUI.

Examples

>>> from numpy.random import randn
>>> fit_data(randn(100),randn(100)) 

This will bring up 100 normally-distributed points with no initial fitting model.

>>> from numpy.random import randn
>>> fit_data(randn(100),randn(100),'linear') 

This will bring up 100 normally-distributed points with a best-fit linear model.

>>> from numpy.random import randn
>>> fit_data(randn(100),randn(100),'linear',weights=rand(100)) 

This will bring up 100 normally-distributed points with a best-fit linear model with the points weighted by uniform random values.

>>> from numpy import tile
>>> from numpy.random import randn,rand
>>> fit_data(randn(100),randn(100),'linear',weights=tile(rand(100),2).reshape((2,10)),fittype='yerr') 

This will bring up 100 normally-distributed points with a linear model with the points weighted by a uniform random number (interpreted as inverse error) fit using the yerr algorithm instead of the default least-squares.

MultiFitGui – Interactive fitting for multiple 1D models

Note that the MultiFitGui requires Mayavi due to the need for 3D plotting.

Reference

class pymodelfit.multifitgui.MultiFitGui(data, names=None, models=None, weights=None, dofits=True, **traits)

Bases: traits.has_traits.HasTraits

data should be c x N where c is the number of data columns/axes and N is the number of points

Parameters:

• data (sequence of c equal-length arrays (length N)) – The data arrays
• names (sequence of strings, length c) – Names
• models (sequence of models, length c-1) – The models to fit for each pair either as strings or astroypsics.models.ParametricModel objects.
• weights (array-like of size N or None) – the weights for each point or None for no weights
• dofits (bool) – If True, the data will be fit to the models when the object is created, otherwise the models will be passed in as-is (or as created).

extra keyword arguments get passed in as new traits (r[finmask],m[finmask],l[finmask]),names=’rh,Mh,Lh’,weights=w[finmask],models=models,dofits=False)

weightsChanged()

pymodelfit.multifitgui.fit_data_multi(data, names=None, weights=None, models=None)

fit a data set consisting of a variety of curves simultaneously. A GUI application instance must already exist (e.g. interactive mode of ipython)

returns a tuple of models e.g. [xvsy,xvsz]