com.numericalmethod.suanshu.stats.regression.linear.modelselection
Class Backward

java.lang.Object
  com.numericalmethod.suanshu.stats.regression.linear.modelselection.Backward

public class Backward
extends java.lang.Object

To construct a GLM model for a set of observations using the backward selection method, we first assume that all factors are included in the model. Iteratively, we drop the least significant factor until all remaining factors are significant.

Nested Class Summary

static class

SingleFactorSelection.ModelNotFound We throw a ModelNotFound exception when we fail to construct a model to explain the data.

Field Summary

double	criticalValue a critical value A factor is considered significant if its z-value is bigger than some critical value.
GlmProblem	problem the GLM problem to be solved

Constructor Summary

Backward(GlmProblem problem, double significance)
Construct automatically a GLM model using the backward selection method.

Method Summary

GeneralizedLinearModel

model() Get the constructed model.

Methods inherited from class java.lang.Object

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

Field Detail

problem

public final GlmProblem problem

the GLM problem to be solved

criticalValue

public final double criticalValue

a critical value

A factor is considered significant if its z-value is bigger than some critical value. We add the significant factors to the model.

Constructor Detail

Backward

public Backward(GlmProblem problem,
                double significance)

Construct automatically a GLM model using the backward selection method.

Parameters:

problem - a GLM problem

significance - a critical value to determine whether a factor is significant (to be included in the model)

Method Detail

model

public GeneralizedLinearModel model()

Get the constructed model.

Returns:

a GLM model