Considering between- and within-person relations in auto-regressive cross-lagged panel models for developmental data

Area of Science:

• Psychometrics
• Quantitative Psychology
• Longitudinal Data Analysis

Background:

• Longitudinal data analysis allows inferences at both between-person and within-person levels.
• Auto-regressive cross-lagged panel models are commonly used for examining time-lagged relations in longitudinal data.
• Existing implementations often fail to adequately distinguish between-person and within-person associations, leading to inaccurate results.

Purpose of the Study:

• To familiarize analysts with common auto-regressive cross-lagged panel model variants.
• To guide researchers in selecting appropriate models based on data characteristics and research questions.
• To improve the accuracy of longitudinal data analysis by addressing model specification issues.

Main Methods:

• Focus on auto-regressive cross-lagged panel models.
• Analysis of common model variants and their interpretations.
• Guidance on selecting models for distinct sources of association in longitudinal data.

Main Results:

• Many common implementations of these models confound between-person and within-person relations.
• Substantial differences in interpretation arise from seemingly minor model specification differences.
• Clearer understanding of how model choices impact results in longitudinal studies.

Conclusions:

• Accurate longitudinal data analysis requires careful selection of statistical models.
• Distinguishing between-person and within-person associations is crucial for valid inferences.
• This work provides guidance for selecting appropriate models to avoid common pitfalls in longitudinal research.