Establishing normative data for repeated cognitive assessment: a comparison of different statistical methods

Area of Science:

• Psychometrics
• Longitudinal Data Analysis
• Neuropsychology

Background:

• Serial cognitive assessment tracks cognitive changes over time.
• Current methods like regression-based change and ANCOVA have limitations, particularly in handling more than two measurement occasions.

Purpose of the Study:

• To propose and evaluate three novel normative methods for serial cognitive assessment.
• To address the limitations of existing methods in analyzing longitudinal cognitive data.

Main Methods:

• Multivariate regression for fixed time points.
• Standard linear mixed model (LMM) for unbalanced data.
• Linear mixed model with multiple imputation (LMM with MI) for handling missing data.

Main Results:

• Multivariate regression is suitable for limited, fixed measurements.
• LMM and LMM with MI are more appropriate for unbalanced, longitudinal data, offering better covariance structure modeling.
• LMM with MI excels at handling missing responses and covariates simultaneously.

Conclusions:

• The proposed multivariate regression, LMM, and LMM with MI methods offer more robust and flexible approaches to serial cognitive assessment.
• These advanced techniques are crucial for accurate normative data development in longitudinal studies.