Paper Summary
Share...

Direct link:

How Much Uncertainty Can Be Acceptable in a Test Specification?

Sat, April 18, 2:45 to 4:15pm, Sheraton, Floor: Second Level, Michigan A

Abstract

Some test specifications define a range for the number of items required in particular content areas. This study examines the impact on test results that occurs when parallel tests are developed based on the same test specification but with different numbers of items across content areas. Preliminary findings indicate that reducing the number of items in content areas with less reliable items slightly increase the reliability of a test with the trade-off of deteriorating values for measure for those content areas and increasing undesired variance in the test results.

Authors