Standard Article

Conditional Standard Errors of Measurement

  1. Ye Tong,
  2. Michael J. Kolen

Published Online: 15 OCT 2005

DOI: 10.1002/0470013192.bsa196

Encyclopedia of Statistics in Behavioral Science

Encyclopedia of Statistics in Behavioral Science

How to Cite

Tong, Y. and Kolen, M. J. 2005. Conditional Standard Errors of Measurement. Encyclopedia of Statistics in Behavioral Science. .

Author Information

  1. University of Iowa, Iowa City, IA, USA

Publication History

  1. Published Online: 15 OCT 2005

Abstract

Errors of measurement for test scores generally are viewed as random and unpredictable. Conditional standard errors of measurement (SEMs) index the amount of error in the measurement process and are required to be reported for tests. Methods of estimating conditional SEMs, based on both classical test theory and generalizability theory, are described in this entry. These methods are based on differing assumptions, but the conditional SEMs estimates obtained from these methods are commonly highly related for educational achievement tests.

Conditional SEMs vary by score level. Conditional SEMs typically have high values for extreme scores and have low values for middle scores on the raw score scale. However, a nonlinear transformation of the raw scores to scale scores can lead to conditional SEMs that have other patterns across score levels.

Keywords:

  • conditional standard errors of measurement;
  • classical test theory;
  • generalizability theory;
  • scale score