Robust Estimation of Ability in the Rasch Model (From Robust Estimation in Latent Trait Analysis - Final Report, 1980 - See NCJ-72747)

• NCJ Number: 72750
• Author(s): H Wainer; B D Wright
• Date Published: 1980
• Length: 43 pages
• Grant Number(s): 78-NI-AX-0047
• Sponsoring Agency: US Dept of Justice
• Corporate Author: Bureau of Social Science Research, Inc
• Sale Source: National Institute of Justice/
• Publication Type: Statistics
• Language: English
• Country: United States of America

Annotation

Formulations aimed at dealing with noise data in estimating ability parameters in the Rasch Model without a reparameterization are tested through a Monte Carlo simulation.

Abstract

Estimating ability parameters in latent trait models in general and in the Rasch Model in particular is almost always hampered by noise in the data. This noise can be caused by guessing, inattention to easy questions, and other factors which are unrelated to ability. It was found that although no one of the tested schemes is uniformly superior to all others, a robustified Jackknife is generally the best scheme; it was also very efficient for tests with 40 or fewer items. The Jackknifing scheme is proposed for estimating ability for practical work. Tabular data and 22 references are provided. (Author abstract modified)