The Center for Social Data Analytics Speaker Series Presents: Siyang Ni (Criminology)

Title: "Is Adolescent Nicotine Vaping Increasingly a Marker of Broader Substance Use?"

Abstract: Youth nicotine vaping in the United States peaked in 2019 and has declined since. But as some teens quit or never start, those who keep vaping may be a different group. Drawing on nationally representative data from more than 50,000 US 12th-grade students in the Monitoring the Future study (2017–2024), I examine whether adolescents who continue to vape are increasingly distinguished by broader substance use. Between 2017–2020 and 2021–2024, alcohol, cannabis, and other non-tobacco substance use came to account for a growing share of what separates students who vape from those who do not, rising from 48% to 58%. Cigarette smoking’s contribution did not change significantly. Cannabis accounted for most of the shift: its use rose among students who vaped and fell among those who did not. This suggests that prevention for teens who still vape should address polysubstance use, not smoking alone. I also take up a broader methodological problem: how to tell whether groups of correlated predictors gain or lose importance over time. Regression coefficients alone can mislead. In this study, smoking’s odds ratio rose sharply, yet its contribution to distinguishing vapers from non-vapers did not grow. I show why both can be true and present a reusable workflow for any repeated cross-sectional data. The workflow combines regularized regression, Shapley decomposition of predictor groups, and cross-period model transport.

Bio: I am a PhD candidate in Criminology and Social Data Analytics at Pennsylvania State University. My research asks how the ways we measure and model development shape what we conclude about how young people grow up and why some get into trouble. I study self-control, delinquency, and substance use from early childhood into adulthood, using large longitudinal and national survey data from the US and UK. I find that seemingly technical decisions about how to analyze data can change which young people we identify as at risk of delinquency, substance use, or later contact with the justice system. I also develop open-source tools so that these methods are more accessible to the research community.