高階相關積之核平滑化無參數

碩士 === 臺中師範學院 === 教育測驗統計研究所 === 91 === The purpose of this research is the development of new item option weight function to improve the estimation of ability proposed by Ramsay. Use the “high level correlation product weighting function” to calculate the score of examine and to estimate the abilit...

Full description

Bibliographic Details
Main Author: 呂雅琇
Other Authors: 劉湘川
Format: Others
Language:zh-TW
Published: 2003
Online Access:http://ndltd.ncl.edu.tw/handle/62042678993458961682
Description
Summary:碩士 === 臺中師範學院 === 教育測驗統計研究所 === 91 === The purpose of this research is the development of new item option weight function to improve the estimation of ability proposed by Ramsay. Use the “high level correlation product weighting function” to calculate the score of examine and to estimate the ability parameter, in order to discuss the accuracy of this model. This research considers about the problem of the evaluation of ability parameter and the limited size of samples. Eventually we decided to use the way of software MATLAB to simulate the data. We try to discuss the result of possible combinations when having different amount of items and examinees especially in the accuracy of the ability estimation analysis. And then compare the “high level correlation product weighting function” and “correlation weighting function” to estimate the accuracy. Mainly in the research we designed 3 different conditions of the tests: 10, 25 and 50 questions; 4 conditions of the examinees: 100, 500, 1000, and 2000 examinees. Hopefully we can understand more by these 12 different combinations. As we discovered by the result of this research, under different combinations of different quantity of test questions and different quantity of examinees, not only the inaccuracy of the estimation decreased while the quantity of the test questions increased, but also the accuracy of “high level correlation product weighting function” become more accurate than the “correlation weighting function”