dwvisser.analysis
Class NonLinearFit

java.lang.Object
  |
  +--dwvisser.analysis.NonLinearFit

Direct Known Subclasses:

GaussianFit

public abstract class NonLinearFit

extends java.lang.Object

This abstract class uses LevenbergMarquadt to do non-linear parametric function fitting. An actual class should define additional parameters and add them to parameters, It should implement estimate(), valueAt(), and derivative().

Version:

0.5, 8/28/98

Author:

Dale Visser

See Also:

valueAt(double), derivative(double, java.lang.String), Fit, LevenbergMarquadt, GaussianFit, Fit#estimate

Field Summary

protected Parameter	chisq the calculated reduced chi-squared statistic
protected double[]	counts The histogram to be fitted.
protected double[]	errors The errors associatied with counts.
protected LevenbergMarquadt	lm does the actual matrix algebra to find the best fit
protected int	maxCH the int value of hi
protected int	minCH the int value of lo

Constructor Summary

NonLinearFit(double[] spectrum, double[] errors, int minCh, int maxCh)
Class constructor.

Method Summary

protected void	addParameter(Parameter newParameter) Makes a parameterTable from parameters.
double	calculate(int channel) Calculate function value for specific channel in the histogram.
abstract double	derivative(double x, java.lang.String parameterName) Evaluates derivative with respect to parameterName at x.
java.lang.String	doFit() Perform fit calulation and return a status String.
abstract void	estimate() Estimate parameters for which the user requests estimation.
Parameter	getParameter(java.lang.String which) Allows access to parameters by name for modification.
double	p(java.lang.String which) Returns double value of parameter indicated by name.
void	setParameter(java.lang.String which, double value) Set a parameter designated by name to a new value.
abstract double	valueAt(double x) Evaluates at x for given parameters.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

counts

protected double[] counts

The histogram to be fitted.

errors

protected double[] errors

The errors associatied with counts.

lm

protected LevenbergMarquadt lm

does the actual matrix algebra to find the best fit

chisq

protected Parameter chisq

the calculated reduced chi-squared statistic

minCH

protected int minCH

the int value of lo

maxCH

protected int maxCH

the int value of hi

Constructor Detail

NonLinearFit

public NonLinearFit(double[] spectrum,
                    double[] errors,
                    int minCh,
                    int maxCh)

Class constructor. This is still an abstract class. Specific subclass constructors will call this before executing their own constructors.

Method Detail

addParameter

protected void addParameter(Parameter newParameter)

Makes a parameterTable from parameters.

See Also:

#parameterTable, #parameters

estimate

public abstract void estimate()
                       throws FitException

Estimate parameters for which the user requests estimation. Unnecessary to implement, but very useful since good guesses are usually needed to find the fit.

Throws:

FitException - thrown if unrecoverable error occurs during estimation

valueAt

public abstract double valueAt(double x)

Evaluates at x for given parameters.

Parameters:

x - value at which to evaluate the fit function

Returns:

value of fit function at x

derivative

public abstract double derivative(double x,
                                  java.lang.String parameterName)

Evaluates derivative with respect to parameterName at x.

Parameters:

parameterName - the name of the parameter to differentiate with respect to

x - value to evalueate at

Returns:

df(x)/d(parameterName) at x

doFit

public java.lang.String doFit()
                       throws FitException

Perform fit calulation and return a status String. Calls LevenbergMarquadt several times, which determines changes in parameter values likely to reduce chi-squared. When reductions are no longer significant, it stops. If there is no covergence, it stops after 10 iterations.

Returns:

message with number of iterations and degrees of freedom to fit

Throws:

FitException - thrown if unrecoverable error occurs during fit

p

public double p(java.lang.String which)

Returns double value of parameter indicated by name.

Parameters:

which - the name of the parameter

Returns:

the current value of the parameter

setParameter

public void setParameter(java.lang.String which,
                         double value)
                  throws FitException

Set a parameter designated by name to a new value.

Parameters:

which - the name of the parameter

value - the value to assign

Throws:

FitException - thrown if unrecoverable error occurs

getParameter

public Parameter getParameter(java.lang.String which)

Allows access to parameters by name for modification.

Parameters:

which - the name of the parameter

Returns:

the Parameter object

calculate

public double calculate(int channel)

Calculate function value for specific channel in the histogram.

Parameters:

channel - the channel to evaluate the function at

Returns:

the value of the function at channel