Scaling MacDonald's AT-20 using item-response theory

The responses of 110 subjects to 18 items of the AT-20 tolerance of ambigui ty scale were analyzed in an item response framework using Rasch modeling. The results indicated that the items satisfy the assumptions of this model and an excellent fit was obtained. In particular, the observed item reliabi lity (KR-20) was 0.93 and the person reliability was 0.68. The latter value agrees with previous findings derived in a classical test theoretical fram ework. A Rasch type scale (M = 25, SD = 5) is developed and various psychom etric and scaling issues are discussed. (C) 1999 Elsevier Science Ltd. All rights reserved.