运用ANN估计小样本的IRT参数研究.pdf

ABSTRACT Item thebasisof itembank．Before so，item responsetheory(IRT)isconstructing doing is mustbe estimated．Onone estimationmethod parameters hand，traditional precisely parameter basedon a mathematical numberofexaminees．Ontheother statistiCS(MS)，whichrequiredlarge numberofexaminees intothe mustbelimitedforthe of hand，the participatingpretest purpose item two ofitemDarameter taskof maketheresearch estimationa security．The veryimportant inthe withsmallnumberofexaminees． IRT,especiallypretest hllllis itern estimationmethodbasedonartificialncnralnetworks thesis．anew parameter orderto its model (ANN)Wasproposed．Inverifyperformance，thethree-parameterLo百stic and neuralnetworkswereselectedastheresearched scoringdichotomouslygeneralregression IRTandANN was andconducted then，simulationexperimentdesigned model，respectively．And toMonteCarlomethod．Conclusionsaredrawnas according foIlowing： An takestheitem IRT as 1． ANN method．which endits modeling response parameter pattern ANN’S and