predictive of alcohol/drug use category membership [X2 (6, 399) = 14.44; p = .025].
Group membership was accurately predicted for 98.1% of the participants based on the
predictors in the model.
Upon further analysis, the primary predictor of category membership was gender
(B = .531, Wald's X2 = 5.18, p = .023). Forty-five percent of the boys were categorized
as having any problems due to alcohol/drug use and only 30% of the girls were. The rest
of the predictor variables (including the interaction term) did not provide a significant
contribution to the model. The second hypothesis was not supported for alcohol/drug use.
A second multiple logistic regression was done without the interaction term
included. The rationale for this test was to create a more parsimonious model to test the
independent contribution of ethnic identity to the prediction of alcohol/drug use category
for all ethnic groups. The overall model maintained its significance with no loss of
accuracy in categorization [X2 (5, 400) = 14.07; p = .015]. The individual predictor
statistics are presented in Table 4.8. This finding demonstrates that ethnic identity does
predict alcohol/drug use for all of the study participants, not simply the minority
adolescents. Those participants with higher ethnic identity scores were more likely to be
in the non-use group than those with lower scores. The mean ethnic identity score for
those in the nonuse group was 3.52. The mean for the use group was 3.33.
Table 4-8. Variable contributions to the prediction of alcohol/drug use category
Variable B SE Wald Sig.
African American -.290 .350 .689 .407
Hispanic .025 .355 .005 .943
European American .283 .364 .604 .437
Gender .526 .233 5.09 .024
MEIM Score* -.507 .202 6.32 .012
*MEIM = Multigroup Ethnic Identity Measure