Content area
Full Text
SUMMARY
A key assumption underlying construct validation is that constructs and their indicators are represented at the appropriate depth (i.e., specific versus general). Bagozzi and Edwards (1998) (hereafter B&E) present a framework that depicts constructs and indicators at various depths and use a formative approach in creating their partial and total aggregation models. Utilizing a scale from organizational research, they conclude that it performs adequately only when used to represent specific work values as opposed to more global, general values. We propose here an alternate method of aggregation. By creating "reflective combinations" of construct indicators, this scale (as well as other multidimensional scales) can indeed represent global, general values.
B&E stress that in any empirical study, it is essential to be specific as to the depth and dimensionality of constructs and their measures if meaningful results are to be obtained (p. 47). They provide a clear example of a multidimensional construct which may be viewed at a general level of abstraction OR at a more precise level, depending on the theoretical model to be studied. The various dimensions of the Work Aspect Preference Scale (WAPS) may be seen in Figure 1. At the most specific level, a researcher might wish to study the constructs of lifestyle, creativity, and self-development and note their impact on other constructs such as job satisfaction or job commitment. Or at the most general level, the researcher might wish to study the impact of work values on job satisfaction and commitment. B&E outline and test several types of aggregated models in order to determine the appropriate levels of abstraction at which the construct of work values might be studied. As their more general models are not supported using SEM, they conclude that work values may only appropriately be studied at its lowest level of abstraction. However, we wish to prove that if modeled correctly, work values can indeed be studied at its more general levels. For brevity purposes, only the partial aggregation model is discussed here.
When using SEM to statistically analyze constructs and the relationships to their indicators as well as to other constructs, the researcher must make use of the reflective measurement model. In order to yield proper results, SEM is predicated on the factor analytic measurement model,...