# Modeling dynamic relations in substance use research: Challenges and opportunities.

> **NIH NIH F31** · UNIV OF NORTH CAROLINA CHAPEL HILL · 2024 · $40,441

## Abstract

PROJECT SUMMARY
Substance use is one of the most commonly occurring health risk behaviors and has been unambiguously
linked to numerous negative outcomes (e.g., Stellern et al., 2023, Planell-Ripoll et al., 2019). These far-
reaching public health issues have impelled substantial growth in the theoretical conceptualizations of causal
processes of substance use, especially in theories of dynamic processes that unfold over time (e.g., Shiffman,
2009). However, the rapid growth in these theories has occurred before the empirical properties of statistical
methodologies used to test the theories have been fully vetted. Multilevel Modeling (MLM) and Dynamic
Structural Equation Modeling (DSEM) are widely used to examine dynamic theories of substance use, yet the
performance of these models under conditions common to theories of substance use is unknown. This directly
undermines the ability of substantive researchers to make informed choices about models. The integration of
theory and methods is essential to building reliable and valid theories of substance use.
MLM is a modeling framework commonly used to assess risk and protective processes of substance use. It
has several strengths including the ability to model unequal time intervals, the use of maximum likelihood
estimation, and the widespread availability of MLM in statistical programs. However, MLM is not well-suited to
model reciprocal relationships (Falkenström et al., 2022), which are hypothesized in many theories of
substance use. Reciprocal effects can be appropriately modeled in DSEM, and DSEM has other strengths
such as the ability to address measurement error and to introduce random effects across a variety of estimates
(Asparouhov et al., 2018). Yet, DSEM is not without limitations when applied to substance use theory. For
example, there is currently no way to systematically build and evaluate DSEM models, which has serious
implications for inference (Hamaker et al., 2018). Further, an assumption of DSEM that is often not met is that
measurements are equally spaced. Undoubtedly, there are conditions in which MLM and DSEM can
adequately and appropriately model dynamic processes of substance use. However, until these conditions are
comprehensively evaluated, we risk building a literature marked by biased results that lack internal validity.
The purpose of this proposed program of research and training is to improve researchers' ability to test novel
theories of substance use, such as the edge on, edge off theory. I aim to (1) generate data consistent with the
characteristics of theories from the substance use literature, (2) identify conditions under which MLM and
DSEM perform optimally and conditions under which they result in biased parameter estimates, and (3)
disseminate guidance for using these methodologies to substance use researchers and meaningfully
contribute to the substance use literature by examining the novel edge-on, edge-off theory. My proposed
project will fully integrate a...

## Key facts

- **NIH application ID:** 10996940
- **Project number:** 1F31DA060040-01A1
- **Recipient organization:** UNIV OF NORTH CAROLINA CHAPEL HILL
- **Principal Investigator:** Jennifer Marie Traver
- **Activity code:** F31 (R01, R21, SBIR, etc.)
- **Funding institute:** NIH
- **Fiscal year:** 2024
- **Award amount:** $40,441
- **Award type:** 1
- **Project period:** 2024-08-01 → 2027-07-31

## Primary source

NIH RePORTER: https://reporter.nih.gov/project-details/10996940

## Citation

> US National Institutes of Health, RePORTER application 10996940, Modeling dynamic relations in substance use research: Challenges and opportunities. (1F31DA060040-01A1). Retrieved via AI Analytics 2026-05-25 from https://api.ai-analytics.org/grant/nih/10996940. Licensed CC0.

---

*[NIH grants dataset](/datasets/nih-grants) · CC0 1.0*