com.itemanalysis.psychometrics.irt.estimation

Interface ItemLogLikelihoodFunction

All Superinterfaces:

DiffFunction, Function, Uncmin_methods

public interface ItemLogLikelihoodFunction
extends DiffFunction, Uncmin_methods

Method Summary

Modifier and Type	Method and Description
double[]	derivativeAt(double[] point) Computes gradientAt of item likelihood function at item parameter values.
int	domainDimension() For DiffFunction interface.
double	f_to_minimize(double[] x) Computes the loglikelihood value for an item.
void	gradient(double[] x, double[] g) Computes the gradientAt of the loglikelihood function for an item.
void	hessian(double[] x, double[][] a) Computes the Hessian of the loglikelihood function for an item.
double	logLikelihood() Item loglikelihood function.
void	setModel(ItemResponseModel model, DistributionApproximation latentDistribution, EstepItemEstimates r, double[] nk)
double	valueAt(double[] point) Computes the loglikelihood value for an item.

Method Detail

setModel

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

logLikelihood

double logLikelihood()

Item loglikelihood function.

Returns:

value of the item loglikelihood function.

domainDimension

int domainDimension()

For DiffFunction interface.

Specified by:

domainDimension in interface Function

Returns:

number of parameters

valueAt

double valueAt(double[] point)

Computes the loglikelihood value for an item. For DiffFunction interface that is used by QNMinimizer in the optimization package.

Specified by:

valueAt in interface Function

Parameters:

point - item parameter values

Returns:

loglikelihood value at item parameter values

derivativeAt

double[] derivativeAt(double[] point)

Computes gradientAt of item likelihood function at item parameter values. For DiffFunction interface that is used by QNMinimizer in the optimization package.

Specified by:

derivativeAt in interface DiffFunction

Parameters:

point - item parameter values.

Returns:

Gradient of item likelihood function at item parameter values

f_to_minimize

double f_to_minimize(double[] x)

Computes the loglikelihood value for an item. For Uncmin_methods interface in the uncmin package. Note that the array indexing begins at 1. For example, a problem with two paramters will require an input array x[] of length = 3. The first index 0 is not used. This indexing is due to the translation of FORTRAN code in the Uncmin optimizer.

Specified by:

f_to_minimize in interface Uncmin_methods

Parameters:

x - input array of item parameters that begins at 1. You must create a new array that begins at 0 and then call valueAt(double[] point) to get the value.

Returns:

Loglikelihood value for an item

gradient

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

Computes the gradientAt of the loglikelihood function for an item. For Uncmin_methods interface in the uncmin package. Note that the array indexing begins at 1. For example, a problem with two paramters will require an input array x[] of length = 3. The first index 0 is not used. This indexing is due to the translation of FORTRAN code in the Uncmin optimizer. If this method is empty, the gradientAt will be computed numerically by the Uncmin optimizer.

Specified by:

gradient in interface Uncmin_methods

Parameters:

x - input array of item parameters that begins at 1. You must create a new array that begins at 0 and then call derivativeAt(double[] point) to get the gradientAt.

g - gradientAt at x.

hessian

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

Computes the Hessian of the loglikelihood function for an item. For Uncmin_methods interface in the uncmin package. Note that the array indexing begins at 1. For example, a problem with two paramters will require an input array x[] of length = 3. The first index 0 is not used. This indexing is due to the translation of FORTRAN code in the Uncmin optimizer. If this method is empty, the Hessian will be computed numerically by the Uncmin optimizer.

Specified by:

hessian in interface Uncmin_methods

Parameters:

x - input array of item parameters that begins at 1. You must create a new array that begins at 0 and then call derivativeAt(double[] point) to get the gradientAt.

a - Hessian at x.