Hilbert, S., Stadler, M., Lindl, A., Naumann, F., & Bühner, M.
Analyzing longitudinal intervention studies with linear mixed models
The predominant majority of researchers conducting randomized control trials or other intervention studies with multiple measurement points still rely on mixed analyses of variance (ANOVAs) or ordinary least squares (OLS) regressions to analyze the differences between their experimental groups. This is problematic as both ANOVA and OLS regressions have several inherent problems, which severely limit the statistical power of these analyses. To mitigate this issue, this paper proposes a linear mixed regression model (LMM) approach to the analysis of differences between multiple groups over multiple time points. After introducing the approach theoretically, its utility is demonstrated via exemplary analyses on two different datasets. The first dataset is used to illustrate the inclusion of continuous covariates, the second one for a comparison with a mixed ANOVA (data and analysis code for both examples are provided in the online repository https://osf.io/k3ph7/). Next to a higher statistical power, the LMM approach provides various methodological advantages over mixed ANOVAs and OLS regressions and is easily adaptable to more complex research designs.
Testing, Psychometrics, Methodology in Applied Psychology, 2019, Vol. 26, pp. 101-119, DOI: 10.4473/TPM26.1.6