Conferences

The primary objective of study 1 was to explore whether the conclusions drawn concerning the psychometric properties of Likert-items and scales depended on whether the classical test theory or the item response theory approaches were used to analyse the data. These techniques differ in terms of the requirements they make in relation to the data analysed. Methods based on CTT presume that the data obtained have metric properties whereas the IRT approach does not make this requirement. Those two models were used to analyse a scale measuring attitude towards homeownership and investigated whether conclusions regarding which items to include in the scale – on the basis of their discrimination power – and reliability depended on which of those techniques were used. In regards to the measurement precision of the scale, the findings showed that the marginal reliability of the IRT analysis was considerably higher than the alpha reliability. Moreover, in regards to the discrimination power of the items, the findings showed that the final scale did depend on which statistical technique was used to analyse the discrimination power of the items. Thus the use of the two methods resulted in scales which defined different constructs. In study 2 two different forms of the home-ownership scale were constructed on the basis of findings from the CTT and IRT analysis. The main purpose of study 2 was to explore whether theory testing was affected by which form of the scale was used to measure the relevant construct. The main findings from this study suggest that the choice of a statistical method to analyse Likert-items and scales may affect the substantial conclusions researchers may draw when testing theories which involve the construct these scales measure.