68

 Standard multiple regression was done with gender, ethnicity, ethnic identity, and

an interaction term predicting the sum of the three mood related symptoms. The

interaction term combined ethnic identity and ethnic minority status. The interaction term

tested the idea that the significant relationship between ethnic identity and mood related

symptoms existed for minority group members only. Gender and ethnicity were dummy

coded and four dichotomous variables were created (African American = 1, Hispanic = 1,

European American = 1; Female = 1). The final model had six predictors. The model was

not predictive of the outcome variable [F (6, 403) = 1.76; p = .106]. Ethnic identity

achievement was not predictive in minorities for this dependent variable. The second

hypothesis was not supported for mood related symptoms.

Ethnic Identity and Alcohol/Drug Use

 The alcohol/drug use subscale was positively skewed with the majority of the

respondents (73.5%) scoring zero on this subscale. Multiple transformations were

unsuccessful in obtaining a normal distribution. Therefore, this measure was changed to a

dichotomous outcome variable (0/1). A score of one indicated a positive response to any

item. This subscale was then used as an outcome in a logistic regression analysis.

 In logistic regression it is assumed not only that the dependent variable is

dichotomous, but also that the options are statistically independent, mutually exclusive,

and collectively exhaustive (Wright, 1997). Also, the model must be adequately specified

(includes all relevant predictors and excludes all irrelevant predictors), and the sample

size must be adequate. Logistic regression requires at least 50 participants per

independent variable (Wright, 1997). All of these assumptions were met in this case.

 Multiple logistic regression was done with the same predictor variables used in

the multiple regression analysis. The results indicated that the overall model was