Navigation

Global	Documentation
Home	Manual
Readme	Glossary
Notes	Design
Examples	Classes
	Trouble
	Style

class MaxLikelihoodFitter( IterativeFitter )	Source

Base class with methods common to fitters handling ErrorDistributions.

Author: Do Kester.

Attributes

• errdis : None | "gauss" | "laplace" | "cauchy" | "poisson" |
                   "uniform" | "exponential"
       None : Use _ChiSq as function to be minimized
       name : use -logLikelihood as function to be minimized from the named
               errordistribution.
• scale : float
       the (fixed) noise scale
• power : float
       power of errdis (if applicable)

Raises

ConvergenceError Something went wrong during the convergence if the fit.

MaxLikelihoodFitter( xdata, model, errdis=None, scale=None, power=2.0, **kwargs )

Create a new iterative fitter, providing xdatas and model.

This is a base class. It collects stuff common to all iterative fitters. It does not work by itself.

Parameters

• xdata : array_like
       array of independent input values

• model : Model
       the model function to be fitted

• errdis : None | "gauss" | "laplace" | "cauchy" | "poisson" |
                   "uniform" | "exponential"
       None : Use _ChiSq as function to be minimized
       name : use -logLikelihood as function to be minimized from the named
               errordistribution.

• scale : float
       the (fixed) noise scale of errdis (if applicable)

• power : float (2.0)
       the power of errdis ( if applicable)

• kwargs : dict
       Possibly includes keywords from
           IterativeFitter : maxIter, tolerance, verbose
           BaseFitter : map, keep, fixedScale

makeFuncs( data, weights=None, index=None, ret=3 )

Make connection to the desired func, gradient and hessian.

Parameters

• data : array_like
       the data to be fitted
• weights : array_like or None
       weights on the data
• index : array_like
       indices of the parameters to be fitted.
• ret : 1 or 2 or 3
       return (func), (func,dfunc) or (func,dfunc,hess)

getScale( )

Return the stdev of the noise.

getLogLikelihood( autoscale=False, var=1.0 )

normalize( normdfdp, normdata, weight=1.0 )

Not Implemented.

Raises

NotImplementedError. the method is not implemented for MaxLikelihoodFitters

testGradient( par, at, data, weights=None )

returns true if the test fails.

Methods inherited from IterativeFitter

• setParameters( params )
• doPlot( param, force=False )
• fitprolog( ydata, weights=None, accuracy=None, keep=None )
• fit( ydata, weights=None, keep=None, **kwargs )
• report( verbose, param, chi, more=None, force=False )

Methods inherited from BaseFitter

• setMinimumScale( scale=0 )
• fitpostscript( ydata, plot=False )
• keepFixed( keep=None )
• insertParameters( fitpar, index=None, into=None )
• modelFit( ydata, weights=None, keep=None )
• limitsFit( ydata, weights=None, keep=None )
• checkNan( ydata, weights=None, accuracy=None )
• getVector( ydata, index=None )
• getHessian( params=None, weights=None, index=None )
• getInverseHessian( params=None, weights=None, index=None )
• getCovarianceMatrix( )
• makeVariance( scale=None )
• getDesign( params=None, xdata=None, index=None )
• chiSquared( ydata, params=None, weights=None )
• getStandardDeviations( )
• monteCarloError( xdata=None, monteCarlo=None)
• getEvidence( limits=None, noiseLimits=None )
• getLogZ( limits=None, noiseLimits=None )
• plotResult( xdata=None, ydata=None, model=None, residuals=True,