AI Insight
This paper argues that evaluation research in fields like clinical intervention, psychotherapy, and organizational change should abandon simple pre-post comparisons in favor of piecewise latent growth modeling (PLGM). The authors contend that traditional difference scores treat change as a single event rather than a continuous, phase-structured process, leading to invalid conclusions. PLGM better captures how interventions unfold over time by modeling distinct phases of change, providing more accurate and interpretable evidence of effectiveness.
Why it matters
Adopting PLGM could fundamentally improve how researchers evaluate interventions across clinical, educational, and organizational settings by revealing not just whether programs work, but how and when change occurs. This shift could lead to better-designed interventions and more nuanced understanding of what drives successful outcomes in real-world applications.
Understand the Science
Evaluation research across applied domains—clinical intervention, psychotherapy, curriculum reform, organizational change—has long relied on the pre–post difference score as its primary evidence of change. This paper argues that this reliance reflects not merely a methodological habit but a systematic ontological error: the implicit treatment of change as a discrete event rather than a continuous, phase-structured process. We develop a theoretical and methodological case for piecewise latent growth modeling (PLGM) as the appropriate evaluative standard across these domains. Drawing on classical psychometric theory, philosophy of measurement, and the longitudinal modeling literature, we establish the conditions under which difference-score logic fails under realistic measurement conditions. We articulate an explicit processual ontology of change and demonstrate its alignment with theoretical commitments already embedded—but rarely operationalized—in leading evaluation frameworks. We present the formal structure of piecewise latent growth models, including two-piece, multi-piece, and growth mixture extensions, illustrated through worked examples from clinical trial evaluation, psychotherapy research, curriculum evaluation in medical education, and organizational change research. We situate PLGM relative to its principal alternatives—polynomial growth models, multilevel models, and exploratory network approaches—and specify the roles of longitudinal measurement-invariance testing and formal model comparison. We address the practical constraints—sample size, measurement frequency, statistical expertise—that limit uptake in resource-constrained settings and propose tiered implementation pathways calibrated to varying analytic capacity. We conclude with a reporting framework for trajectory-based evaluation studies. The argument rests on a single claim: evaluation research cannot produce interpretively valid evidence of change until it adopts methods whose ontological commitments match the phenomena it studies. The field must transition from asking “Did it work?” to “How did it unfold?”—because only the latter produces evidence worthy of its name.
Source: Piecewise latent growth modeling as the methodological standard for evaluation research