Longitudinal Data Analysis Using Mixed Effects Models

Longitudinal data analysis using mixed effects models offers a flexible framework to investigate change over time while accounting for correlation among repeated measures on the same units. By partitioning variation into fixed effects, which capture population-level trends, and random effects, which reflect individual-level deviations in intercepts and slopes, these models accommodate unbalanced designs, missing data under mild assumptions, irregular measurement intervals and hierarchical structures. Estimation typically employs maximum likelihood or restricted maximum likelihood to derive unbiased parameter estimates and standard errors, with Bayesian methods gaining traction for complex or sparse datasets.

Applications span diverse fields including clinical trials, developmental neuroscience, ecology and environmental monitoring, where time-varying covariates and nonlinear growth patterns are common. Extensions such as piecewise growth models address shifts in trajectories at known change-points, latent factor mixed models integrate dimensionality reduction for high-dimensional endpoints, and mixture formulations uncover latent subgroups with distinct developmental pathways. Model selection and diagnostics draw on information criteria, likelihood ratio tests and visualisation of residual patterns to ensure theoretical coherence and empirical adequacy.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Research from all publishers

Recent work provides a beginner-friendly guide to selecting among multilevel and latent curve frameworks, emphasising practical implementation in open-source software and clarifying when mixed effects specifications better capture complex within- and between-subject dynamics.

A novel multivariate framework combines probabilistic principal component analysis with longitudinal linear mixed effects modelling to characterise treatment response in clinical trials of psoriatic and rheumatoid arthritis, handling missing data and correlational structure to derive latent trajectories that explain over 70 percent of empirical variation.

Power analysis for piecewise linear growth under attrition has been advanced by embedding a survival model for dropout into a multilevel mixed model, revealing how timing of turning points and attrition patterns influence required sample sizes and guiding researchers in study design optimisation.

Longitudinal Data Analysis Using Mixed Effects Models publication trend

The graph below shows the total number of articles in longitudinal data analysis using mixed effects models across all publications each year (not limited to Nature Index journals).

Technical terms

Mixed effects model: A statistical model that includes both fixed effects (common to all units) and random effects (unit-specific deviations) to capture hierarchical or repeated-measures structure.

Fixed effect: A parameter representing the average influence of a covariate on the response variable across the entire population.

Random effect: A latent variable capturing variation in intercepts or slopes across individual units or clusters.

Restricted maximum likelihood (REML): An estimation technique that adjusts for the loss of degrees of freedom when estimating variance components, yielding less biased variance estimates.

Piecewise growth model: A longitudinal model in which the trajectory is divided into segments with distinct linear slopes joined at specified change-points.

Latent trajectory: An unobserved, smooth path of change over time for an individual, inferred by the mixed effects model’s random components.

References

1. The Hitchhiker’s guide to longitudinal models: A primer on model selection for repeated-measures methods. Developmental Cognitive Neuroscience (2023).
2. A framework for longitudinal latent factor modelling of treatment response in clinical trials with applications to Psoriatic Arthritis and Rheumatoid Arthritis. Journal of Biomedical Informatics (2024).
3. Power analysis of longitudinal studies with piecewise linear growth and attrition. Behavior Research Methods (2022).