Estimating Relative Stability in Developmental Research: A Critique of Modern Approaches and a Novel Method

• Published in: Journal of quantitative criminology 2017-06, Vol.33 (2), p.319-346
• Main Authors: Barnes, J. C., El Sayed, Sarah A., TenEyck, Michael, Nedelec, Joseph L., Connolly, Eric J., Schwartz, Joseph A., Boutwell, Brian B., Wright, John P., Beaver, Kevin M., Anderson, Nathaniel E.
• Format: Article
• Language: English
• Subjects: Criminologists; Criminology; Criminology and Criminal Justice; Deviance; Law and Criminology; Life course; Life events; Methodology of the Social Sciences; Methods; ORIGINAL PAPER; Researchers; Sociology; Statistics; Studies
• ISSN: 0748-4518; 1573-7799
• DOI: 10.1007/s10940-016-9298-5

Objective Developmental/life-course (DLC) criminologists often study the age-graded trajectories of traits and behaviors known to correlate with antisocial outcomes. Much of this work has attempted to discern whether traits like impulse control are relatively stable across different portions of the life course. A range of statistical techniques have been employed by researchers attempting to parameterize relative stability. Yet, despite these attempts, much of the evidence remains mixed. Methods We draw on data from the Pathways to Desistance study to examine whether the methods typically used to analyze longitudinal development provide a parameter estimate for relative stability. Results The results of our demonstration reveal that none of the methods typically employed by DLC researchers provide a parameter estimate for relative stability. In order to address this oversight, we develop a novel method—P(Δ)—that can be used to estimate the amount of relative (in)stability that is observed in a longitudinal dataset. Conclusions Although P(Δ) provides a direct estimate of the degree to which relative (in)stability is observed in one’s dataset, there are several important points that must be considered by future DLC researchers in order to further develop P(Δ) into a statistic that can be used for inferential analysis. We consider these points in the discussion.