Christophersen Knut-Andreas, Helseth Sølvi, Lund Thorleif
Institute of Political Science, University of Oslo, Box 1097, Blindern, 0381, Oslo, Norway.
Qual Life Res. 2008 Feb;17(1):87-93. doi: 10.1007/s11136-007-9289-y. Epub 2007 Dec 11.
The aim in the present study was to illustrate generalizability theory by exploring some psychometric aspects of the Norwegian version of KINDL (KINDL-N). This questionnaire is a measure of quality of life in adolescents. It consists of 24 items arranged in six subscales or aspects, corresponding to six domains of adolescents' health-related quality of life.
A total of 239 healthy adolescents completed the questionnaire. The analysis was performed in line with generalization theory, and this analysis yields among other things so-called generalizability and dependability coefficients, which are analogous to reliability coefficients in classical test theory. The preferred design was a mixed two-facet design, with the six aspects as a fixed facet and the items nested within aspects as a random facet. For comparison, a random two-facet design and a one-facet design were also analyzed.
The generalizability and dependability coefficients for KINDL-N are satisfactory, but the coefficients can be markedly improved by increasing the length of the instrument while keeping it to a manageable size. The mixed two-facet design resulted in somewhat higher coefficients than a one-facet design and much higher coefficients than a random two-facet design.
If the items within a questionnaire are arranged in distinct subscales, a two-facet design analyzed by generalizability theory gives more sophisticated results than classical test theory. Moreover, the results may depend considerably on whether a mixed or random two-facet design is chosen. Hence, the choice of an appropriate measurement design in line with the intended use of the questionnaire is essential.
Zotero 插件