Bias and Interpretation: Cases for Ordinal Measurement

Author(s):

Myers, Charles T.

Publication Year:

1973

Report Number:

RM-73-18

Source:

ETS Research Memorandum

Document Type:

Report

Page Count:

9

Subject/Key Words:

Grades (Scholastic), Mathematical Models, Predictive Measurement, Statistical Bias, Test Bias

Abstract

High school and college grades are suggested to have the nature of ordinal scales rather than interval scales; hence the median is the appropriate and preferred average, rather than the mean. The use of a median grade point average has some of the major advantages of pass-fail grading. Assuming grades to be ordinal rather than interval data suggests that estimates of test bias should not be subjected only to analysis of covariance. As an alternative, equipercentile equating procedures are suggested as a method of studying bias that would tend to be more favorable for low-scoring groups.

Read More

• Request Copy (specify title and report number, if any)