Reading tips on longitudinal mixed models and mediation

Question

I'm a Swedish PhD student in psychology looking for advice on what to read to understand linear mixed models in a longitudinal context better. We're running a RCT with two active groups and weekly measures that will continue to collect data for a couple of years. When it's over the intention is to investigate a couple of putative mediators in this data. Our analysis plan includes using linear mixed models for this purpose, but I'm early in my career and so far only have a surface level understanding of the method. Since the data collection will not be completed until end of next year, I hopefully have time to learn it.

I would be happy to receive recommendations on what to read to get a deeper level of understanding here, both theoretical and practical. Books, articles, online accessible courses, youtube lectures, anything is welcome.

My background knowledge:

• Somewhat proficient in R.
• In my opinion a good enough understanding of basic ideas in frequentist null-hypothesis testing and basic methods in inferential statistics.
• Good enough understanding of simple and multiple regression.
• However, weak background in math. Would not be able to derive or prove anything I'm using from first principles but can usually follow along with (good) explanations of it.

Answer 1

For a good overview of issues in longitudinal data analysis, I recommend Chapter 7 of Frank Harrell's Regression Modeling Strategies. Mixed models aren't the only, or even necessarily the best, way to analyze data from studies like yours. There's a helpful table summarizing the strengths and weaknesses of several types of approaches, and references for further reading. Its emphasis is on generalized least squares rather than on mixed models, but I think it's important to understand what you gain and what you lose if you jump immediately to linear mixed models without considering alternatives.

Answer 2

I like the somewhat old book by Hedeker and Gibbons: Longitudinal Data Analysis and also Gelman and Hill Data Analysis and Multilevel Models. Frank Harrell's book (recommended by EdM) is excellent, as well, and gives you much more than just material on mixed models.

Answer 3

Analyzing mediation, under the broader heading of "causal inference," can be tricky. This answer by Robert Long is a superb, concise introduction to the directed acyclic graphs that help clarify your understanding of the subject matter in a way that informs your modeling. For a remarkably helpful introduction to causal inference in general, read Causal Inference: What If by Hernán and Robins. It develops the essential concepts through simple examples "without models" and builds from there through increasingly complicated scenarios. Its 13 pages of references provide many resources for more detailed study.