Classical Test Theory and Criterion-Referenced Scales

The item (difficulty and discrimination) and test (reliability and validity) statistics in classical test theory are highly dependent upon the calibration sample of individuals used. The estimates of item and test parameters in classical test theory is valid within a range of interest along the char...

Full description

Saved in:
Bibliographic Details
Main Author Woodson, M. I. Charles E
Format Publication
LanguageEnglish
Subjects
Classical Test Theory
Criterion Referenced Tests
Item Analysis
Research Reports
Statistics
Test Construction
Test Reliability
Test Validity
Online AccessGet full text

Cover

More Information
Summary:The item (difficulty and discrimination) and test (reliability and validity) statistics in classical test theory are highly dependent upon the calibration sample of individuals used. The estimates of item and test parameters in classical test theory is valid within a range of interest along the characteristic measured. Generally, this range of interest is the distribution of the characteristic in some population, and the calibration sample used is intended to be a random sample from that population. In such populations, the extremes usually are poorly represented, and the parameter estimates are relatively poor at these extremes. For criterion-referenced scales, the range of interest is defined by a range of the characteristic rather than the distribution of that characteristic in some population. The calibration sample must be representative of that range of interest. When the range of interest is appropriately defined, an appropriate calibration sample may be selected, and classical test theory applies directly to criterion-referenced scales. (Author/DB)