Date of Award
Doctor of Philosophy
Underage drinking has featured prominently in both scholarly and conventional literature in recent decades as a major health and socio-economic concern in the United States. As new evidence emerges associating underage drinking with a host of negative outcomes for both the youth who drinks and society in general, a closer examination of the long-term effects of underage drinking is critical. This exploratory study was designed to examine predictor variables and their ramifications (1) using logistic regression to identify a model for underage drinking history (UDHISTORY) as a predictor of concomitant alcohol dependence and poor educational attainment (CADAPEA) among individuals aged 25 and above, and (2) obtain a better understanding of how demographic variables (age, gender, race/ethnicity) influence the prediction. The nature and strength of the effect(s) of these demographic variables on the prediction were also investigated. The 2010 National Survey on Drug Use and Health data set ICPSR 32722-0001 which is previously unexploited for this purpose is utilized in this study. The data analysis tool, SDA on SAMHSA's website and IBM SPSS were used for correlation analysis and logistic regression to test the hypothesis that currently legal age drinkers 25 years and older with UDHISTORY are more likely to experience CADAPEA than their counterparts without UDHISTORY. When considered alone, UDHISTORY was a strong and statistically significant predictor of CADAPEA. The identified bivariate logistic regression model was statistically significant, &chi2 (1, n = 60) = 13.39, Adjusted Wald F1, 60 = 13.39, p = 0.001 < .05, accounting for 1.26% (Cox and Snell R square), 1.3% (Log Likelihood Pseudo R square), to 7.9% (Nagelkerke R square) of the variance in CADAPEA. However, adding demographic variables to the model made UDHISTORY a much stronger and more statistically significant predictor. The identified final multivariable logistic regression model was statistically significant, &chi2 (6, n = 55) = 170.43, Adjusted Wald F6, 55 = 26.04, p = 0.00 < .001, accounting for 1.8% (Cox and Snell R square), 7.2% (Log Likelihood Pseudo R square) to 7.9% (Nagelkerke R square) of the variance in CADAPEA. The model also correctly classified 99.1% of cases.
This dissertation is Open Access and may be downloaded by anyone.