Journal of Statistics: Advances in Theory and ApplicationsAuthor's: Yi Pan, Qiu Wang and Michael Thomas Willoughby
Pages: [135] - [157]
Received Date: March 28, 2015
Submitted by:
DOI: http://dx.doi.org/10.18642/jsata_7100121479
Researchers in the social and behavioural sciences often ask questions regarding the time-linked associations between two or more constructs at multiple assessments across time. Although their questions typically involve an interest in making within-subject inferences, the most commonly used analytic approach does not disaggregate between-from within-subject variations, resulting in a mismatch between question and method. In this paper, we compare and contrast three analytic approaches that can be used to test time-linked (typically cross-lagged) associations between two variables that are repeatedly measured over time. Through use of a real-world example and Monte Carlo simulation, we demonstrate how two of these approaches facilitate within-subject inferences. We also provide practical advice to researchers to help determine which analytic approach is most appropriate for their research question.
autoregressive model, autoregressive latent trajectory model, cross-lagged model, fixed influence cross-lagged model, repeated measures analysis, structural equation model.
