Latent growth curve modeling in exercise science

ABSTRACT

Longitudinal measurement enables the examination of behavioral or psychological change. One approach to examining longitudinal measurements is the use of latent growth curve modeling (LGCM). This approach affords the assessment of inter- and intraindividual change. Yet, this approach likely is underused in exercise science. The purpose of the current study was to describe and demonstrate the use of LGCM to examine change using multiple measurements in the field of exercise science. We first provide a substantive review of LGCM. We highlight the use of unconditional models to find an appropriate model of change, how and why to utilize autoregressions, and how to examine predictors of change in conditional models. We then provide an illustration of the approach using data from the Michigan State Motor Performance Study. In the conclusion, we discuss the advantages and limitations of the approach and suggest future directions when assessing longitudinal data in exercise science.

KEYWORDS:

• Autoregression
• growth curves
• longitudinal analysis
• motor performance

Supplementary material

Supplemental data for this article can be accessed here.

Notes

¹ A quadratic model is not fit in the illustration used. Readers can find an illustration of this model in Ram and Grimm (2007).

² The intercept and slope of a LGCM can also serve as predictors of outcomes (e.g., change in physical activity level could predict health status). Another possibility is to fit two growth curves in the same model and then correlate the intercept and slope factors. Readers can refer to Grimm et al. (2017) as well as Hancock and Mueller (2013) for more detail on model extensions with predictors.