On the reliability of didactic tests

Authors:

Pages: 242-248

Abstract:

Didactic tests should be reliable and valid. The reliability is the self-validity of a given measurement tool therefore there is no doubt that it is of a main importance in any didactic measurement. On the other hand the reliability represents the correlation between the test “observed score” and the test “true score” of the construct under measurement. Given a test with a poor reliability is useless. One can investigate reliability by means of various classical reliability indexes like the Kronbach alpha or the Guttman split-half reliability coefficient. Modern test theory (Item Response Theory) offers some probability models which purpose in fact is to achieve the test reliability to some extent. The most common used model is the 3-parametric Birnbaum model in the case of a test with multiple-choice item type. The most simple 1-parametric model (the Rasch model) is also used in the case of a test with free-answer items type. Here we discuss some practical rules by help of which the teacher can verify the reliability of a given test. This question arises always when the teacher manufactures some achievement or criterion test for its own use. Another situation, when the reliability is of main importance, is the case when the teacher want to use given test which is standardized with respect to different conditions or the test is designed for different population.

Keywords:

Download

NA / Algeria / Bulgaria / China / Cote D'Ivoire / Europe / France / Germany / Portugal / Russian Federation / Singapore / Turkey / Ukraine / United Kingdom / United States