Data Analyses
Preliminary analyses examined descriptive statistics and
bivariate correlations among study variables. Subsequent
analyses used SEM procedures and the statistical software
Mplus (Muthe´n and Muthe´n 2001). In general, models with
CFI values[.90 and RMSEA values.10 are considered
adequate fits to the data (Hu and Bentler 1999). We tested
the structural model hypothesized in Fig. 1 twice: once for
internalizing symptoms and once for externalizing symptoms.
Parent and child reports of symptoms were included
in the same models. We initially controlled for youth age in
the full models, but age was not significantly associated
with any variables, and thus, final analyses did not control
for age. Preliminary analyses comparing models for boys
and girls suggested that model fits did not differ by gender;
therefore, we did not constrain models by gender in the
final analyses. As displayed in Fig. 1, we tested the full
models examining the effects of SES and parent positive
communication on follow-up symptoms while accounting
for baseline symptoms.