g., bisexual women, gay men). Consequently, our second set of hypotheses focus on questions of heterogeneity within each sexual orientation category. To those 17-DMAG side effects ends, we stratified analyses by gender and sexual orientation group, which allowed us to test research questions such as ��Do gay men who experience discrimination have increased odds of smoking when compared with their gay male counterparts who do not experience discrimination?�� Methods This project used the National College Health Assessment (NCHA) data, collected by the American College Health Association. Each fall and spring semester, post-secondary educational institutions choose to purchase and administer the NCHA with their students.

To form national datasets, the American College Health Association (2009c) combines data from institutions that used either a random or a census sample design and gained institutional review board approval to implement the survey. As a secondary analysis of de-identified data, the authors�� institutional review board considered this project nonhuman subjects research. From the total combined sample of 113,790 participants, we first selected only those respondents of 18�C24 years old (n = 92,470) in order to comprise a young adult-only sample (Park, Mulye, Adams, Brindis, & Irwin, 2006). Sexual orientation was measured with an item asking respondents to self-identify as heterosexual (n = 85,710), gay/lesbian (n = 1,825), bisexual (n = 2,545), or unsure (n = 1,545). For the purposes of this project, we defined sexual minority as persons who identified as lesbian, gay, or bisexual.

Persons who were unsure of their sexual orientation were analyzed as a separate group. Since analyses were stratified by sexual orientation and heterosexuals comprised over 90% of the sample, to keep the heterosexual analyses from being statistically overpowered, a random 5% subsample was drawn from the original heterosexual young adult sample (n = 4,286). Using chi-square tests of independence, comparisons were made between the random 5% heterosexual subsample and the original heterosexual sample on demographics (i.e., sex, race, and age) and found no statistically significant differences (data not shown). Consequently, the final analytic sample for all analyses was 11,046. Participating institutions administered the survey via either the Internet or paper-and-pencil.

Overall, paper-and-pencil surveys generated higher mean response rates than Web-based formats (Paper2008 = 63%; Web2008 = 22%; Paper2009 = 82%; Web2009 = 20%). Globally, the mean response rates among participating institutions in the Fall 2008 and Spring 2009 semesters were 27% and 30%, respectively (American College Health Association, 2009a, AV-951 2009b). Detailed analyses of the NCHA II survey reliability and validity are available from the American College Health Association (n.d.).