Concepts (T)

Longitudinal Data Analysis

The section offers a brief overview of longitudinal data analysis, focusing on key concepts and methods. It explains that longitudinal data involves the repeated measurement of variables of interest over time, often referred to as panel data. The example study data is initially in wide-form, with multiple variables for each time point, and then it is reshaped into long-form data for analysis.

The analysis of longitudinal data includes various models, such as linear models, mixed-effect models, and marginal models (GEE). Linear models are introduced initially, where each subject has a common slope and intercept. These ideas can be expanded to incorporate random intercepts, random slopes, and combinations of both in mixed-effect models. Model diagnostics, including AIC, BIC, and -loglik, are discussed for model evaluation. The section briefly discussing different correlation structures and highlights the differences in interpretation between mixed models and marginal models. This section also includes an introduction to marginal structual model, which is a GEE model under a situation when treatment-confounder feedback exists.

Reading list

Key reference: (Faraway 2016; Hothorn and Everitt 2014)

Optional references: (Karim et al. 2021; Cui and Qian 2007)

Video Lessons

Longitudinal data formatting
Longitudinal models: mixed effect
Longitudinal models: GEE
Marginal structural model

Video Lesson Slides

Longitudinal models

Marginal structural model

References

Cui, Jianwen, and Guoqing Qian. 2007. “Selection of Working Correlation Structure and Best Model in GEE Analyses of Longitudinal Data.” Communications in Statistics—Simulation and Computation® 36 (5): 987–96.
Faraway, Julian J. 2016. Extending the linear model with R: generalized linear, mixed effects and nonparametric regression models. CRC press.
Hothorn, Torsten, and Brian S Everitt. 2014. A Handbook of Statistical Analyses Using r. CRC press.
Karim, Mohammad Ehsanul, Helen Tremlett, Feng Zhu, John Petkau, and Elaine Kingwell. 2021. “Dealing with Treatment-Confounder Feedback and Sparse Follow-up in Longitudinal Studies: Application of a Marginal Structural Model in a Multiple Sclerosis Cohort.” American Journal of Epidemiology 190 (5): 908–17.