

“Ethnic Variations in Characteristics of First Unions” 191
African American women (Schoen, Landale, Daniels, & Cheng, 2009).
A four-category maternal education variable was created: less than
high school, high school graduates, some college, and university and
above. Less than high school is used as the reference group.
Generational status is undoubtedly important in immigrant studies and
thus is also included in the analyses. A three-category variable was
created, with the third-generation as reference group.
D. Missing Values and Complex Survey Design
There are few missing cases in the analytical samples. For race,
maternal education, and generational status, the percentages of
missing cases are about 1% or fewer. Listwise deletion was used to
handle missing data. The complex survey design of the Add Health
study was also taken into account during model estimations.
Appropriate longitudinal sampling weights were applied to the
statistical models. In addition, clustering and stratifying variables
were both applied to adjust the standard errors.
E. Statistical Analyses
Descriptive statistics are first presented to offer an overview
of the study population and to point out general patterns of
differences in union characteristics in each racial/ethnic group.
Statistics are shown for all race/ethnicity groups as well as for all
Asian subgroups. Next, logistic regression models were fit to
explore the differences in first union behaviors, such as the
likelihood of forming a cohabiting and a marital union by Wave 4,
the outcome of a first cohabitation and whether a first marriage is
preceded by a pre-marital cohabitation. Finally, the race of the first
cohabiting/marital partner by racial/ethnic group was also
examined. Logistic regression models were first fit for all
respondents and then for Asian Americans only to reveal subgroup
differences as shown in Tables 2 and 3.