natural mentor and adverse mental health outcomes (i.e., marijuana use) and
demonstrated that those adolescents that had a natural mentor had significantly lower use
[y = (4.71-3.90)/4.34 = 0.18] than those that did not have a mentor (Zimmerman et al.,
2002). A difference of this size would be considered clinically relevant on the outcome
measures for this study.
Table 3-1. Statistical analyses
Research Aims Hypotheses Analysis
I. Evaluate the I. Those adolescents who I. Multivariate Analysis of
relationship between have natural mentors Variance (MANOVA)
the presence of a have better mental
mentor and mental health functioning than
health functioning of those who do not have
the mentee. natural mentors.
II. Examine the II. There is a relationship II. Multiple Linear
relationship between between ethnic identity Regression & Logistic
the ethnic identity of and mental health Regression.
the adolescent and functioning in minority
his or her mental adolescents.
health functioning.
III. Evaluate the III. The presence of III. Multiple Linear
potential moderating ethnically matched Regression was used to
effect of an ethnic mentor moderates the test for a significant
mentor on ethnic relationship between interaction.
identity of the ethnic identity and
mentee and his or mental health
her mental health outcomes in minority
functioning. adolescents.
A simple analysis for a dichotomous predictor variable (natural mentor, yes or no)
and a continuous outcome variable (alcohol/drug use subscale) was utilized to estimate
sample size. Assuming an alpha (two-tailed) = .05, and a beta of 0.20 (power = 0.80),
how many participants are needed to demonstrate a 6.25% difference on the alcohol/drug
use subscale, a small effect size? Utilizing an estimated standardized effect size of 0.20
SD, approximately 197 participants are needed per group to avoid a Type II error with