Statistical analyses
Consistent with literature cited above, we expected that
parental monitoring would have a protective effect on
high school drinking and that high school drinking would
correlate strongly with college drinking. We further
hypothesized that parental monitoring would have an
indirect protective effect on college drinking via its influence
on high school drinking. To test these hypotheses,
we performed a series of ordinary least squares multiple
regressions predicting (1) high school drinking on the
basis of parental monitoring, (2) college drinking on the
basis of parental monitoring, and (3) college drinking on
the basis of both parental monitoring and high school
drinking. For each analysis, hypothesized effects were first
evaluated at the bivariate level as were the effects of several
control variables (race, sex, mother's education, religiosity,
current living situation, time in college, and combined
SAT score). Next, all explanatory variables were evaluated
together in a series of multiple regression models. In the
final model, the first-order interaction term of parental
monitoring with high school drinking was included to
test for a possible moderating effect. Because students
were interviewed at various times during their first year in
college, the combined models included time in college
(i.e., the number of months from arrival on campus to the
interview date) to adjust for any potential confounding
effects related to the timing of the interview. Effect size for
each explanatory variable was evaluated using the semipartial
r2 (sr2) statistic, which represents the proportion of
variance in the outcome variable uniquely explained by
an explanatory variable, while holding constant the
remaining explanatory variables in the model
Statistical analysesConsistent with literature cited above, we expected thatparental monitoring would have a protective effect onhigh school drinking and that high school drinking wouldcorrelate strongly with college drinking. We furtherhypothesized that parental monitoring would have anindirect protective effect on college drinking via its influenceon high school drinking. To test these hypotheses,we performed a series of ordinary least squares multipleregressions predicting (1) high school drinking on thebasis of parental monitoring, (2) college drinking on thebasis of parental monitoring, and (3) college drinking onthe basis of both parental monitoring and high schooldrinking. For each analysis, hypothesized effects were firstevaluated at the bivariate level as were the effects of severalcontrol variables (race, sex, mother's education, religiosity,current living situation, time in college, and combinedSAT score). Next, all explanatory variables were evaluatedtogether in a series of multiple regression models. In thefinal model, the first-order interaction term of parentalmonitoring with high school drinking was included totest for a possible moderating effect. Because studentswere interviewed at various times during their first year incollege, the combined models included time in college(i.e., the number of months from arrival on campus to theinterview date) to adjust for any potential confoundingeffects related to the timing of the interview. Effect size foreach explanatory variable was evaluated using the semipartialr2 (sr2) statistic, which represents the proportion ofvariance in the outcome variable uniquely explained byan explanatory variable, while holding constant theremaining explanatory variables in the model
การแปล กรุณารอสักครู่..
