HOW TO SHORTEN A TEST AND INCREASE ITS RELIABILITY AND VALIDITY
Abstract
Logic and mathematics are employed to yield very conservative estimates of the gains resulting from changing over from choice methods to admissible probability measurement in the administration of existing tests. Equations and graphs give test reliability and measurement validity as a function of the distribution of ability levels in the population to be tested and as a function of the amount and type of guessing engaged in by this population. Since guessing degrades the performance of choice tests and since the use of admissible probability measurement eliminates guessing, the extent of degradation corresponds to a conservative estimate of the gains resulting from conversion to admissible probability measurement. In some applications it may be wise to trade off the increase in measurement validity against the advantages of shortening the length of the test. Equations and graphs show how much shorter the new guessing-free test can be and still remain the original measurement validity. Additional equations and curves show that choice tests with zero measurement validities can have appreciable reliabilities due to differences in guessing strategy in the population. All the analyses indicate that conversion to admissible probability measurement will yield quite significant improvements in measurement validity along with considerable reductions in test length.
Document Details
- Document Type
- Technical Report
- Publication Date
- May 01, 1967
- Accession Number
- AD0677207
Entities
People
- Emir H. Shuford Jr.