Invalid measurement of constructs in survey research often remains undetected and can lead to false conclusions. An important determinant of a construct’s measurement validity is how it is modeled. A construct can often be modeled in different ways, such as the sum of its parts or the cause of its effects. Since each of these models is associated with a unique set of errors, the common practice of specifying only a single model undermines validity. Current guidelines on measurement have not focused on how better validity can be achieved by comparing and combining multiple models. In this paper we provide a framework for the development and use of multiple models. This, we hope, would lead researchers view their construct of interest from different perspectives and thus measure it more validly.
Zwanenburg, Sander Paul, "Valid Construct Measurement Using Multiple Models" (2015). ECIS 2015 Research-in-Progress Papers. Paper 8.