com.itemanalysis.psychometrics.irt.estimation

Class EstepParallel

java.lang.Object

java.util.concurrent.ForkJoinTask<V>

java.util.concurrent.RecursiveTask<EstepEstimates>

com.itemanalysis.psychometrics.irt.estimation.EstepParallel

All Implemented Interfaces:

java.io.Serializable, java.util.concurrent.Future<EstepEstimates>

public class EstepParallel
extends java.util.concurrent.RecursiveTask<EstepEstimates>

Estep of the EM algorithm for estimating item parameters in MMLE. The computation is done in parallel if the number of response vectors exceeds estepParallelThreshold.

See Also:

Serialized Form

Constructor Summary

Constructors

Constructor and Description
EstepParallel(ItemResponseVector[] responseVector, ItemResponseModel[] irm, DistributionApproximation latentDistribution, int start, int length) Default constructor may be called recursively for parallel computations.

Method Summary

Modifier and Type	Method and Description
protected EstepEstimates	compute() Recusive computation of the Estep for parallel processing.
protected EstepEstimates	computeDirectly() Estep computation when the number of response vectors is less than estepParallelThreshold or when the recursive algorithm has reached its stopping condition.

Methods inherited from class java.util.concurrent.RecursiveTask

exec, getRawResult, setRawResult

Methods inherited from class java.util.concurrent.ForkJoinTask

adapt, adapt, adapt, cancel, compareAndSetForkJoinTaskTag, complete, completeExceptionally, fork, get, get, getException, getForkJoinTaskTag, getPool, getQueuedTaskCount, getSurplusQueuedTaskCount, helpQuiesce, inForkJoinPool, invoke, invokeAll, invokeAll, invokeAll, isCancelled, isCompletedAbnormally, isCompletedNormally, isDone, join, peekNextLocalTask, pollNextLocalTask, pollTask, quietlyComplete, quietlyInvoke, quietlyJoin, reinitialize, setForkJoinTaskTag, tryUnfork

Methods inherited from class java.lang.Object

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

Constructor Detail

EstepParallel

public EstepParallel(ItemResponseVector[] responseVector,
                     ItemResponseModel[] irm,
                     DistributionApproximation latentDistribution,
                     int start,
                     int length)

Default constructor may be called recursively for parallel computations.

Parameters:

responseVector - response vectors for a given set of data

irm - an array of item response models whose parameters are being estimated

latentDistribution - latent distribution quadrature used for computing the marginal likelihood

start - beginning index for the response vector. Manual calls should always be 0, recursive calls are done automatically.

length - length of the response vector segment. This length is manually set to the length of the response vector. Recursive calls use the value specified by estepParallelThreshold.

Method Detail

computeDirectly

protected EstepEstimates computeDirectly()

Estep computation when the number of response vectors is less than estepParallelThreshold or when the recursive algorithm has reached its stopping condition. In parallel processing, this method represents each chunk that is processed in parallel.

Returns:

compute

protected EstepEstimates compute()

Recusive computation of the Estep for parallel processing. It will aggregate all chunks that are processed in parallel.

Specified by:

compute in class java.util.concurrent.RecursiveTask<EstepEstimates>

Returns:

values of nk, rjk, and the loglikelihood (without priors).