com.itemanalysis.psychometrics.irt.estimation

Class ItemLogLikelihood

java.lang.Object

com.itemanalysis.psychometrics.irt.estimation.ItemLogLikelihood

All Implemented Interfaces:

DiffFunction, Function, Uncmin_methods

public class ItemLogLikelihood
extends java.lang.Object
implements DiffFunction, Uncmin_methods

This class contains contains methods for the loglikelihood function for an item, which is the value of the objective function that is being minimized in the Mstep to obtain the maximum likelihood or Bayes modal estimates. See MstepParallel. This class is also used in computation of the starting values. See StartingValues. It applies to binary and polytomous item response models.

Constructor Summary

Constructors

Constructor and Description
ItemLogLikelihood()

Method Summary

Modifier and Type	Method and Description
double[]	derivativeAt(double[] point) Method required for DiffFunction interface to QNMinimizer See ItemDichotomous.h in ETIRM
int	domainDimension() Method required for DiffFunction interface to QNMinimizer
double	f_to_minimize(double[] x) Method required by Uncmin_methods interface
void	gradient(double[] x, double[] g) Method required by Uncmin_methods interface
void	hessian(double[] x, double[][] a) Method required by Uncmin_methods interface
double	logLikelihood(double[] iparam) Log-likelihood for an item.
void	setModel(ItemResponseModel model, DistributionApproximation latentDistribution, EstepItemEstimates r, double[] nk)
double[]	stdError(double[] x, double[][] hessian) Computes the standard errors of the item parameter estimates.
void	stdError(ItemResponseModel irm)
double	valueAt(double[] point) Method required for DiffFunction interface to QNMinimizer

Methods inherited from class java.lang.Object

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

Constructor Detail

ItemLogLikelihood

public ItemLogLikelihood()

Method Detail

setModel

public void setModel(ItemResponseModel model,
                     DistributionApproximation latentDistribution,
                     EstepItemEstimates r,
                     double[] nk)

logLikelihood

public double logLikelihood(double[] iparam)

Log-likelihood for an item.

Returns:

log-likelihood

domainDimension

public int domainDimension()

Method required for DiffFunction interface to QNMinimizer

Specified by:

domainDimension in interface Function

Returns:

valueAt

public double valueAt(double[] point)

Method required for DiffFunction interface to QNMinimizer

Specified by:

valueAt in interface Function

Parameters:

point - parameter values

Returns:

value of objective function

derivativeAt

public double[] derivativeAt(double[] point)

Method required for DiffFunction interface to QNMinimizer See ItemDichotomous.h in ETIRM

Specified by:

derivativeAt in interface DiffFunction

Parameters:

point - parameter values. The order of the parameters must match the order in the gradientAt function in the ItemResponseModel class.

Returns:

gradientAt values

f_to_minimize

public double f_to_minimize(double[] x)

Method required by Uncmin_methods interface

Specified by:

f_to_minimize in interface Uncmin_methods

Parameters:

x - parameter values

Returns:

gradient

public void gradient(double[] x,
                     double[] g)

Method required by Uncmin_methods interface

Specified by:

gradient in interface Uncmin_methods

Parameters:

x - parameter values

g - gradientAt values

hessian

public void hessian(double[] x,
                    double[][] a)

Method required by Uncmin_methods interface

Specified by:

hessian in interface Uncmin_methods

Parameters:

x - parameter values

a - hessian matrix

stdError

public double[] stdError(double[] x,
                         double[][] hessian)

Computes the standard errors of the item parameter estimates.

Parameters:

x - array of item parameter estimates

Returns:

array of item parameter standard errors

stdError

public void stdError(ItemResponseModel irm)