Abstract
Studies that include multiple assessments of a particular instrument within the same population are based on the presumption that this instrument measures the same construct over time. But what if the meaning of the construct changes over time due to one's experiences? For example, the experience of a traumatic event can influence one's view of the world, others, and self, and may disrupt the stability of a questionnaire measuring posttraumatic stress symptoms (i.e., it may affect the interpretation of items). Nevertheless, assessments before and after such a traumatic event are crucial to study longitudinal development of posttraumatic stress symptoms. In this study, we examined measurement invariance of posttraumatic stress symptoms in a sample of Dutch soldiers before and after they went on deployment to Afghanistan (N = 249). Results showed that the underlying measurement model before deployment was different from the measurement model after deployment due to invariant item thresholds. These results were replicated in a sample of soldiers deployed to Iraq (N = 305). Since the lack of measurement invariance was due to instability of the majority of the items, it seems reasonable to conclude that the underlying construct of PSS is unstable over time if war-zone related traumatic events occur in between measurements. From a statistical point of view, the scores over time cannot be compared when there is a lack of measurement invariance. The main message of this paper is that researchers working with posttraumatic stress questionnaires in longitudinal studies should not take measurement invariance for granted, but should use pre- and post-symptom scores as different constructs for each time point in the analysis.
Highlights
Questionnaires are often used at different time points to assess mean or individual change over time
A questionnaire to assess posttraumatic stress symptoms can be rated at different time points after a traumatic event to study the course of problematic responses
Statisticians have stressed the importance of testing measurement invariance when comparing latent mean scores over time (e.g., Byrne et al, 1989; Steenkamp and Baumgartner, 1998; Vandenberg and Lance, 2000), the assumption that factor loadings and intercepts of the underlying items are equal over time often seems to be taken for granted
Summary
Questionnaires are often used at different time points to assess mean or individual change over time. Statisticians have stressed the importance of testing measurement invariance when comparing latent mean scores over time (e.g., Byrne et al, 1989; Steenkamp and Baumgartner, 1998; Vandenberg and Lance, 2000), the assumption that factor loadings and intercepts (or thresholds when dealing with dichotomous or categorical scores instead of continuous scores) of the underlying items are equal over time often seems to be taken for granted. In this article we discuss a measure that, from a theoretical perspective, is expected to lack measurement invariance In such cases the solutions of establishing partial invariance (Byrne et al, 1989) or approximate invariance (van de Schoot et al, 2013; Muthén, 2014) are not a valid solution. We will test for measurement invariance in two samples, and investigate causes of measurement non-invariance and interpretations of the results in this situation
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
