shortening of test, abbreviated version, reliability
Creating shortened versions of research tools is common and justified. Unfortunately, it is often performed without due methodological care and awareness of the consequences of such actions. Even though the errors committed during short form construction were collected by Smith and his collaborators in 2000, it did not distinctly affect the practice. The mistakes made by researchers still come down to two main faults: assuming the transferability of validity and reliability between the full and shortened versions, and lowering the validity and reliability requirements for short forms. These two problems manifest as 9 sins committed during the construction of short forms. This article intends to present procedures which prevent these mistakes and ensure creating possibly the most reliable short version of a research tool and assessing the costs of a selected shortening method. To this end, the work determined a priori the expected length of the tool, the benefit of reduced questionnaire completion time in relation to the cost of reliability loss. Also, it estimated overlapping variance of the full and short version and classification accuracy of the new, short version. Since there are many statistical techniques of questionnaire shortening, an additional effect of this article is a comparison of the efficiency of shortening by means of three various techniques. The results show similarity between the method based on factor loadings and Cronbach’s α method, and a slight advantage over the two of a method based on difficulty and discriminatory power in the IRT paradigm.
Paweł Kleka, Adam Mickiewicz University in Poznan
Institute of Psychology
Emilia Soroko, Adami Mickiewicz University in Poznan