An item selection procedure to maximise scale reliability and validity

Abstract

Wille (1996) proposed an item selection strategy which may be used to maximise, first, the internal consistency and, next, the convergent and discriminant validity of items in multi-dimensional Likert-type questionnaires or scales. In terms of his strategy, the latter aspects of validity are maximised by means of exploratory factor analyses. In this article, it is done by means of Tateneni, Mels, Cudeck and Browne’s (2001) Comprehensive Exploratory Factor Analysis (CEFA) program which implements exploratory factor analysis, but provides the advantages of standard confirmatory factor analysis (e.g., the computation of the standard errors of the rotated factor loadings and measures of “model" fit). The benefits that accrue by using this incremental approach are demonstrated in terms of Allport and Ross’ (1967) Religious Orientation Scale, a widely-used psychological instrument. Opsomming Wille (1996) het ’n itemseleksiestrategie voorgestel om eerstens die interne konsekwentheid, en tweedens die konvergente en divergente geldigheid van items in multidimensionele Likert-tipe vraelyste of skale te maksimeer. Volgens sy strategie word laasgenoemde aspekte van geldigheid deur middel van eksploratiewe faktorontledings gemaksimeer. In hierdie artikel, sal dit gedoen word deur Tateneni, Mels, Cudeck en Browne (2001) se program vir Omvattende Eksploratiewe Faktorontleding (CEFA) te gebruik, wat eksploratiewe faktorontleding aanwend, maar ook die voordele van gewone, bevestigende faktorontleding (bv., die berekening van die standaardfoute van die geroteerde faktorbeladings en indekse van modelpassing) bied. Die voordele wat spruit uit die toepassing van hierdie inkrementele benadering word gedemonstreer aan die hand van Allport en Ross (1967) se Religious Orientation Scale, ’n gewilde sielkundige meetintrument.