TPC Journal-Vol 10- Issue 2-FULL ISSUE

160 The Professional Counselor | Volume 10, Issue 2 increasingly complex social networks (Arnett, 2004), social resources are conceptualized as a third level of constructs in the present model, after internal resources and emerging adulthood identification. Present Study The present study was designed to address several literature gaps concerning college student mental health and well-being. First, it combines several disparate threads of related research by testing a model including internal resources (i.e., attachment security and ego resilience), identification with the dimensions of emerging adulthood, and social resources (i.e., social support and social media usage). Although some research has examined the additive impact of more than one of these sets of constructs together (e.g., attachment and social support), no existing research has examined all three collectively. Second, the present study examined the mental health implications of emerging adulthood and social media usage: two constructs that are the result of 21st century societal forces. A primary hypothesis of the study was that each predictor variable set would explain unique and additive variance for two characteristics of college student mental health (i.e., psychological well-being [PWB] and life satisfaction). A secondary hypothesis was that emerging adulthood identification and social media usage would predict unique variance in each outcome variable even after accounting for the effects of all other predictor variables in the model. Method Participants and Procedure Participants in this IRB-approved study were traditional-aged undergraduate students from a large, public university in a metropolitan area of the Pacific Northwest. Participants were recruited via a recruitment email sent to a random sample of students meeting the inclusion criteria (i.e., 18 to 25 years old and enrolled as a full-time undergraduate student). An a priori power analysis was conducted to determine appropriate sample size (Faul et al., 2007). Given the large number of variables in the model and the fact that Hypothesis 2 was based on semipartial correlations, a small-to-medium effect size was selected ( f 2 = .08). Results suggested an ideal sample size of approximately 400 participants. Assuming an approximate 10% response rate (Manfreda et al., 2008), recruitment emails were sent to 4,000 undergraduates. The recruitment email contained a link to an online survey containing all demographic and study variable items. Surveys were received from 616 undergraduates (15.4% response rate). Data were treated according to the recommendations for multivariate analysis by Meyers et al. (2013). That is, 56 cases (9.1%) were removed because they contained missing data on at least 50% of the items. An additional 17 cases (2.8%) were removed for indicating that they were no longer paying attention at the midpoint of the survey. The remaining missing values were replaced with their respective itemmean because no item was missing more than seven cases (1.3%) and no variable contained more than two missing items for any remaining participant. Data were screened for multivariate outliers using Mahalanobis distance, resulting in the removal of five (0.9%) participants. Thus, the study sample consisted of 538 participants. The study sample had a mean age of 21.72 years ( SD = 2.05) and was predominantly female ( n = 378, 70.3%), while other participants identified as male ( n = 142, 26.4%) or other ( n = 16, 3.0%), and two participants declined to answer. The sample was racially diverse, as 341 (63.4%) participants identified as White, 64 (11.9%) as Latinx, 63 (11.7%) as Asian or Pacific Islander, 14 (2.6%) as Black or African American, 11 (2.0%) as Arab American or Middle Eastern, eight (1.5%) as Native American, 27 (5.0%) as multiracial, and seven (1.3%) as other, while three participants declined to answer.