Reliability and Validity of Test Instruments

Topics Covered in this Session

Standardized Test

Standardized tests have been used for decades by educational researchers in many countries. They are usually developed by experts and are generally well constructed. Individual test items are analyzed and revised until they meet standards of quality. Objectivity is the goal of most standardized tests that is, they are not biased against individuals based on their traits. In addition standardized tests are expected to be reliable and valid. Reliability refers to consistency of of results. Validity refers to what a test measures and for whom it is appropriate. Examples of standardized tests include the SAT, GRE, Degrees of Reading Power, Piers-Harris, etc.

Reliability and Validity

Correlation coefficients are used extensively to measure reliability (consistency of results) and validity (measures what it is supposed to measure) of standardized tests. Simple reliability tests usually involve giving the same test to the same sample of subjects two or more times and then comparing results using correlations which would expected to be very high. Another approach involves having the same sample of subjects take different questions of the same test (i.e. odd numbered questions versus even numbered questions) and compare results using correlation coefficients. Booklets that accompany standardized tests will generally refer to or identify their reliability coefficients. Validity tests for content of subject matter are usually done by a panel of experts and not by using statistical analyses. Validity tests for making predictions such as the use of SAT scores to predict college GPA., usually use correlation coefficients of the two measures to establish their predictive validity. In establishing reliability and validity of standardized tests, very high correlations (i.e. +.80 and above) are generally expected.