Using Causal Graphs to Explain Confounding Bias From Individual Differences in Cross-Lagged Panel Models

Researchers use cross-lagged panel models to study how variables are associated with each other and causally affect one another over time. However, observed associations can reflect both causal effects and stable differences between individuals, potentially producing confounding bias in estimates of population-average causal effects. Accordingly, models such as the random-intercept cross-lagged panel model statistically separate stable between-person differences from within-person variation. To clarify the causal interpretation of these models, previous work used linear structural causal models and causal graphs to make the required causal assumptions explicit. We extend this work to nonlinear parametric structural causal models that represent individual differences in intercepts, causal effects, and innovation variances through underlying causes that enter the causal mechanisms of psychological processes additively and through interactions with lagged variables and innovations, generating differences in person-specific trajectories and fluctuations around them. Using examples, we illustrate how unobserved person-specific causes open backdoor paths and how their interactions can confound estimates of population-average causal effects even after accounting for additive between-person differences. In this way, structural causal models and their implied causal graphs provide a framework for explaining individual differences in psychological dynamics in terms of explicit causal assumptions about how these differences arise, giving researchers a basis for deciding which varying parameters and dependencies to estimate in statistical panel models and for assessing whether the resulting coefficients can be interpreted as estimates of population-average causal effects.

Authors

Institutions

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-18
DOI
https://doi.org/10.5281/zenodo.24338019
Primary Topic
Advanced Causal Inference Techniques
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Using Causal Graphs to Explain Confounding Bias From Individual Differences in Cross-Lagged Panel Models

Moritz Ketzer, Manuel C. Voelkle
1 links
Zenodo (CERN European Organization for Nuclear Research)
Advanced Causal Inference Techniques
preprint

Using Causal Graphs to Explain Confounding Bias From Individual Differences in Cross-Lagged Panel Models

Moritz Ketzer, Manuel C. Voelkle
preprint en

Abstract

Researchers use cross-lagged panel models to study how variables are associated with each other and causally affect one another over time. However, observed associations can reflect both causal effects and stable differences between individuals, potentially producing confounding bias in estimates of population-average causal effects. Accordingly, models such as the random-intercept cross-lagged panel model statistically separate stable between-person differences from within-person variation. To clarify the causal interpretation of these models, previous work used linear structural causal models and causal graphs to make the required causal assumptions explicit. We extend this work to nonlinear parametric structural causal models that represent individual differences in intercepts, causal effects, and innovation variances through underlying causes that enter the causal mechanisms of psychological processes additively and through interactions with lagged variables and innovations, generating differences in person-specific trajectories and fluctuations around them. Using examples, we illustrate how unobserved person-specific causes open backdoor paths and how their interactions can confound estimates of population-average causal effects even after accounting for additive between-person differences. In this way, structural causal models and their implied causal graphs provide a framework for explaining individual differences in psychological dynamics in terms of explicit causal assumptions about how these differences arise, giving researchers a basis for deciding which varying parameters and dependencies to estimate in statistical panel models and for assessing whether the resulting coefficients can be interpreted as estimates of population-average causal effects.

Zenodo (CERN European Organization for Nuclear Research)
University of Freiburg (DE), Humboldt-Universität zu Berlin (DE), Max Planck Institute for Human Development (DE)
Advanced Causal Inference Techniques
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon