Abstract
Objective
We evaluated how applying post-stratification sampling weights to National College Health Assessment II (NCHA-II) data affects estimates of substance use prevalence and tests of medical and recreational marijuana legalization (MML and RML) effects. Participants/Methods: We constructed weights for Fall 2015 and Spring 2016 surveys (n = 90,503) using population information on U.S. undergraduates’ gender and race/ethnicity and three institutional characteristics (region, city population, public/private). We estimated substance use prevalence (e.g., e-cigarettes, prescription opioid misuse) and compared 30-day marijuana use rates in states with RML, MML, or neither policy. Results: When unweighted versus weighted data were used, prevalence estimates did not differ appreciably; conclusions from logistic regressions were similar (weighted 30-day marijuana use rates among undergraduates in RML, MML, and non-ML states were 30.0%, 20.3%, and 16.3%, respectively) but effect sizes differed. Conclusions: The value of using weighted NCHA-II data depends on the analysis and the precision required for the research questions.
Acknowledgments
The authors thank Mary Hoban and Christine Kukich for their consultation, expertise, and assistance. The opinions, findings, and conclusions presented/reported in this article/presentation are those of the authors, and are in no way meant to represent the corporate opinions, views, or policies of the American College Health Association (ACHA). ACHA does not warrant nor assume any liability or responsibility for the accuracy, completeness, or usefulness of any information presented in this article/presentation.
Conflict of interest disclosure
The authors have no conflicts of interest to report. The authors confirm that the research presented in this article met the ethical guidelines, including adherence to the legal requirements, of the United States and was deemed exempt by the Oregon State University Institutional Review Board.
Notes
1 Unweighted and weighted estimates (standard errors) were 0.286 (0.0042) and 0.300 (0.0048) for RML, 0.196 (0.0018) and 0.203 (0.0028) for MML, and 0.152 (0.0021) and 0.163 (0.0033) for non-ML states.