Edith Cowan University
Place of Publication
Perth, Western Australia
Measurement Assessment and Evaluation Laboratory
This article examines the precision of measurements obtained from using the Rasch Dichotomous Model to analyse test data. Considering tests in which the item difficulties are uniformly spaced from easiest to most difficult, permits the derivation of an alternative expression for the standard error of measurement. This expression is sufficiently simple to enable the precision properties of uniform tests to be readily described and to enable a variety of problems of test construction to be solved. One particular problem is that of best test design. Regarding measurement precision as a property of the test only, we show that the best uniform test of a given length and a given target interval is the one that satisfies a minimax condition on the standard error. We illustrate the solution to this problem and describe properties of best tests.