Beyond Pre-Post Comparisons: A Comprehensive Effect Size Framework for Intensive Longitudinal Interventions via Time-Varying Modeling
Abstract
Intensive longitudinal interventions (ILIs), which deliver repeated interventions in people’s daily lives while intensively tracking their psychological states, offer unique opportunities for ecologically valid, real-time, and personalized support. However, there is a striking mismatch between the richness of the data these studies generate and the simplicity of the methods used to evaluate them: evaluations still rely largely on pre-post mean comparisons that collapse dozens or even hundreds of observations per person into a few summary statistics. To resolve this mismatch, we introduce a comprehensive effect size framework based on Bayesian multilevel time-varying autoregressive (TV-AR) models within the generalized additive mixed model (GAMM) approach. The framework organizes effect sizes along four dimensions: subject level (population-average, individual-specific, and between-group), parameter type (mean level, autoregressive effect, and variance), temporal scale (integrated and time-specific), and study phase (active intervention and post-intervention decay). These metrics move researchers beyond whether an intervention works to when effects emerge, plateau, or fade, how the intervention reshapes the dynamics and stability of a process, and for whom it is effective. We demonstrate the framework with data from a mindfulness-based ILI targeting negative affect, in which an intervention and a control group were assessed before and during the intervention, and provide openly available annotated R code and a tutorial. Overall, the framework provides a practical and flexible toolkit for evaluating ILIs, supporting the broader shift toward a more dynamic, process-oriented, and personalized intervention science.
Related
- Intensive Interventions for Burnout, Stress and Occupational Well-Being: A Critical Review and a Delivery-Format Framework
- Outcomes of Comprehensive Multiple Sclerosis Care: A Pilot Comparison of Intensive, Integrated, and Usual Care Approaches
- Reducing Noise and Alarm Burden in Adult Intensive Care Units: An Integrative Review of Interventions and Implications for Nursing Management
- Comprehensive Echocardiographic Assessment of Right Ventricular Afterload and Right Ventricular-Pulmonary Arterial Coupling in the Intensive Care Unit.
- Clinical Profile, Indications, and Maternal Outcomes of Obstetric Intensive Care Unit (ICU) Admissions: A Retrospective Study With Emphasis on Critical Care Interventions.
BibTeX
@misc{luo2026beyond,
title = {{Beyond Pre-Post Comparisons: A Comprehensive Effect Size Framework for Intensive Longitudinal Interventions via Time-Varying Modeling}},
author = {Xiaohui Luo and Yue Liu and Hongyun Liu and Laura Bringmann},
howpublished = {PsyArXiv},
year = {2026},
month = sep,
doi = {10.31234/osf.io/tm6k8_v1},
url = {https://osf.io/preprints/psyarxiv/tm6k8_v1/},
note = {CC0},
}
RIS
TY - UNPB
TI - Beyond Pre-Post Comparisons: A Comprehensive Effect Size Framework for Intensive Longitudinal Interventions via Time-Varying Modeling
AU - Xiaohui Luo
AU - Yue Liu
AU - Hongyun Liu
AU - Laura Bringmann
PY - 2026
DA - 2026/09/16
PB - PsyArXiv
DO - 10.31234/osf.io/tm6k8_v1
UR - https://osf.io/preprints/psyarxiv/tm6k8_v1/
AB - Intensive longitudinal interventions (ILIs), which deliver repeated interventions in people’s daily lives while intensively tracking their psychological states, offer unique opportunities for ecologically valid, real-time, and personalized support. However, there is a striking mismatch between the richness of the data these studies generate and the simplicity of the methods used to evaluate them: evaluations still rely largely on pre-post mean comparisons that collapse dozens or even hundreds of observations per person into a few summary statistics. To resolve this mismatch, we introduce a comprehensive effect size framework based on Bayesian multilevel time-varying autoregressive (TV-AR) models within the generalized additive mixed model (GAMM) approach. The framework organizes effect sizes along four dimensions: subject level (population-average, individual-specific, and between-group), parameter type (mean level, autoregressive effect, and variance), temporal scale (integrated and time-specific), and study phase (active intervention and post-intervention decay). These metrics move researchers beyond whether an intervention works to when effects emerge, plateau, or fade, how the intervention reshapes the dynamics and stability of a process, and for whom it is effective. We demonstrate the framework with data from a mindfulness-based ILI targeting negative affect, in which an intervention and a control group were assessed before and during the intervention, and provide openly available annotated R code and a tutorial. Overall, the framework provides a practical and flexible toolkit for evaluating ILIs, supporting the broader shift toward a more dynamic, process-oriented, and personalized intervention science.
ER -