改進值譜數之測驗能力值估算新模式

A Novel Algorithm to Estimate Examinee's Ability with Greater Spectrum Number

10.7108/PT.200412.0233

陳桂霞(Guey-Shya Chen)；鄭裕篤(Yu-Du Cheng)；劉湘川(Hsiang-Chuan Liu)

倒傳遞類神經網路 ； 試題反應理論 ； 點二相關係數鑑別指數 ； 多元計分模式 ； 值譜數 ； artificial neural network, back-propagation network ； item response theory ； point-biserial correlation item discrimination index ； multivariate logistic categorical response model ； spectrum number

測驗學刊

51卷2期（2004 / 12 / 01）

233 - 248

繁體中文

本研究將「倒傳遞類神經網路」結合「多元計分模式」與「點二系列相關鑑別指數」運用在測驗能力估算上，提供一個新的測驗能力估算模式。再以電腦模擬大量的施測資料，針對試題數及樣本人數進行分析印證，並與BILOG-MG軟體作能力估算的精準度比較，類神經網路的估算模式能得到更精準的測驗能力估計值，尤其當題目數愈少時兩者差距愈顯著。運用多元計分模式，能區分同分考生之能力高低，且在能力值估算上本模式比BILOG-MG有較多的值譜數。

This research tries to combine the model of back-propagation neural network with the multivariate logistic categorical response model and point-biserial correlation item discrimination index to obtain more accurate estimated values of examinee's ability. We do some computer simulations for various item numbers and sample sizes to verify better performance of our proposed model by comparing the results estimated by BILOG-MG. And from the simulation results we also observe that the proposed approach has greater spectrum number than that of BILOG-MG.

1. Mislevy, R. J.,Bock, R. D.(1990).BILOG 3 : Item analysis and test scoring with binary logistic models.
2. Muraki, E.,Bock, R. D.(1998).PARSCALE (version 3.5). Parameter scaling of rating data.
3. Thissen, D.,Chen, W. H.,Bock, R.D.(2003).Multilog (version 7) [Computer software].
4. White, H.(1988).IEEE international Joint Conference on Neural Networks.
5. Wingersky, M.,Patrick, R.,Lord, F. M.(1988).LOGIST.
6. Yen, W. M.(1987).Pschometrika, Vol.52.
7. Zimowski, M. F.,Muraki, E.,Mislevy, R. J.,Bock, R. D.(1996).BILOG-MG: Multiple-Group IRT Analysis and Test Maintenance for Binary Items.
8. 何榮桂(1999)。量身訂製的測驗-適性測驗。測驗與輔導，153(12)
9. 陳桂霞、鄭裕篤、劉湘川(2003)。倒傳遞類神經網路在測驗能力值估算之模擬研究。
10. 劉湘川(2000)。多元計分三參數試題選項分析固定效應模式。測驗統計年刊，8
11. 劉湘川(2000)。點二系列相關試題鑑別指數之值譜分析及其在IRT上之應用。測驗統計年刊，8