Leveraging Intensive Longitudinal Data in Social Work Research: A Gentle Introduction to Two-Level Dynamic Structural Equation Modeling

Many areas of social work research benefit from the analysis of longitudinal data, which enable researchers to model developmental processes and their associations with other variables of interest. In many cases, longitudinal studies aim to test theories and hypotheses about long-term developmental processes experienced by respondents, with measurement occasions spread months (or even years) apart. Data collection schedules of this sort are appropriate for modeling developmental processes that are expected to unfold over relatively long periods of time. These types of data can be analyzed from a variety of analytic frameworks, including multilevel modeling (e.g., random coefficients growth modeling) and structural equation modeling (e.g., latent growth curve modeling, cross-lagged panel modeling).

Importantly, it is possible for relatively long-term and dispersed longitudinal data collection and analysis approaches to gloss over complexity and nuance inherent in the more granular minutia of respondents' experiences. Indeed, in many areas of inquiry germane to social work there remain valuable opportunities to explore relatively short-term longitudinal processes that can be observed on a granular level, such as hour-to-hour or day-to-day, which are features of intensive longitudinal data (ILD). Methodological approaches such as daily diaries, experience sampling, ambulatory assessment, and ecological momentary assessment are intended to collect data daily, hourly, or even more intensively. Thus, ILD enables assessments of how variable values fluctuate around a respondent's baseline level during the study window, and whether theoretically plausible correlates meaningfully predict such fluctuation; that is, the goals of research leveraging ILD is often to understand (a) why there are "peaks and valleys" across short periods of time in a key variable of interest and (b) whether such dynamics vary significantly across respondents (McNeish & Hamaker, 2020).

Two-level dynamic structural equation modeling (DSEM) is particularly well equipped to analyze ILD, as it effectively integrates the analytic advantages of (a) time-series analysis, (b) multilevel modeling, and (c) structural equation modeling. Together, these features enable the nuanced assessment of within-unit longitudinal processes (level-1 dynamics) as well as the assessment of between-unit variability in those within-unit processes (level-2 dynamics). The core aims of this workshop are to highlight differences between long-term longitudinal data collection efforts and the collection of short-term ILD; describe how ILD can be used to answer novel and interesting research questions related to various social work topics; and gently overview powerful analytic strategies, such as two-level DSEM, that are well positioned to analyze ILD.

Workshop attendees will be engaged in both didactic and interactive content delivery processes. The didactic component will be used to provide user-friendly overviews of (a) ILD and how it differs from common longitudinal data collection efforts, (b) strategies for collecting ILD, (c) how ILD could enrich various research areas relevant to social work, and (d) two-level DSEM and how it can be used efficiently in statistical software such as Mplus. The interactive component of the workshop will include embedded opportunities for attendees to engage with the material being presented, ask questions, and discuss how the material could be applied directly to their own areas of research.