A comparison of mixed-models for the analysis of non-linear longitudinal data: application to late-life cognitive trajectories

Maude Wagner,Donald R. Hedeker,Tianhao Wang,Graciela Muniz-Terrera,Ana W. Capuano

2024-03-06

Abstract:Several mixed-effects models for longitudinal data have been proposed to accommodate the non-linearity of late-life cognitive trajectories and assess the putative influence of covariates on it. No prior research provides a side-by-side examination of these models to offer guidance on their proper application and interpretation. In this work, we examined five statistical approaches previously used to answer research questions related to non-linear changes in cognitive aging: the linear mixed model (LMM) with a quadratic term, LMM with splines, the functional mixed model, the piecewise linear mixed model, and the sigmoidal mixed model. We first theoretically describe the models. Next, using data from two prospective cohorts with annual cognitive testing, we compared the interpretation of the models by investigating associations of education on cognitive change before death. Lastly, we performed a simulation study to empirically evaluate the models and provide practical recommendations. Except for the LMM-quadratic, the fit of all models was generally adequate to capture non-linearity of cognitive change and models were relatively robust. Although spline-based models have no interpretable nonlinearity parameters, their convergence was easier to achieve, and they allow graphical interpretation. In contrast, piecewise and sigmoidal models, with interpretable non-linear parameters, may require more data to achieve convergence.

Methodology

What problem does this paper attempt to address?

The problems that this paper attempts to solve are: in the non - linear changes of the cognitive trajectories of the elderly, the performance and applicability of different mixed - effect models. Specifically, the author compared five mixed models for analyzing non - linear longitudinal data to evaluate the performance of these models in describing cognitive trajectories in later life and to provide researchers with guidance on selecting and interpreting these models.

### Research Background and Problems
As people age, changes in cognitive ability are usually non - linear. Especially near the end of life, cognitive ability may decline at an accelerated rate. In order to better understand this non - linear change, researchers have proposed a variety of mixed - effect models to capture these complex dynamic processes. However, there is still a lack of systematic comparison of these models, especially regarding the effectiveness of their performance and interpretation in practical applications.

### Specific Problems
1. **Model Selection**: Which models are most suitable for describing the non - linear changes in cognitive trajectories in later life?
2. **Model Interpretation**: How do different models interpret the influencing factors of cognitive change (such as education level)?
3. **Model Performance**: How do different models perform in fitting non - linear changes? Can they accurately capture the key features of cognitive change?

### Solutions
The author solved the above problems through the following steps:
1. **Theoretical Description**: The mathematical forms and characteristics of the five mixed models are introduced in detail.
2. **Empirical Analysis**: Using the data of two prospective cohorts, the performance of these models in explaining the impact of education on cognitive change was compared.
3. **Simulation Study**: Simulation experiments were carried out to evaluate the performance of different models in various situations and to provide practical suggestions.

### Model Comparison
The five models compared in the paper include:
- **Linear Mixed Model (LMM) with Quadratic Term**
- **Linear Mixed Model (LMM) with Spline Function**
- **Functional Mixed Model (FMM)**
- **Piecewise Linear Mixed Model (PMM) with Random Change Point**
- **S - shaped Mixed Model (SMM)**

Each model has its own unique structure and application scenarios. By comparing these models, the author hopes to provide a more definite selection guide for future research.