Measurement Error in Nonparametric Item Response Curve Estimation. Research Report. ETS RR-11-28.
To read the full text of this article, click here: http://www.eric.ed.gov/ERICWebPortal/contentdelivery/servlet/ERICServlet?accno=ED523692
Nonparametric, or kernel, estimation of item response curve (IRC) is a concern theoretically and operationally. Accuracy of this estimation, often used in item analysis in testing programs, is biased when the observed scores are used as the regressor because the observed scores are contaminated by measurement error. In this study, we investigate the deconvolution kernel estimation of IRC, correcting for the measurement error in the regressor variable. Using item response theory (IRT) simulated data and some real data, we compared the traditional kernel estimation and the deconvolution estimation of IRC. Results show that in capturing important features of the IRC, the traditional kernel estimation is comparable to the deconvolution kernel estimation in item analysis. Measurement Error is appended. (Contains 3 tables and 10 figures.)
|Printer friendly Cite/link Email Feedback|
|Author:||Guo, Hongwen; Sinharay, Sandip|
|Date:||Apr 3, 2010|
|Previous Article:||Content and Language Integrated Learning in Higher Technical Education Using the "InGenio" Online Multimedia Authoring Tool.|
|Next Article:||Practical Use of ICT in Science and Mathematics Teachers' Training at Dar es Salaam University College of Education: An Analysis of Prospective...|