[Using multivariate multilevel model in studying the influential factors of violence among rural high school students...

To provide evidence for setting up violence intervention programs in rural middle schools, through studying the influential factors.

Taking variables including emotional, physical and sexual violence in the past year as the multi-dependent variables before multivariate multilevel model logistic regression model was adopted to analyze the correlations among the three kinds of violence and the influential factors.

Among 3620 respondents, the incidence rates of emotional, physical and sexual violence weres 21.5%, 24.3% and 2.0% respectively. The correlation coefficients between emotional violence vs. physical violence, emotional violence vs. sexual violence, physical violence vs. sexual violence were 0.337, 0.133, 0.131 respectively when the random effect of class difference was separated by multivariate multilevel model. There was an internal aggregation of the incidence rate on physical violence in different grades (chi2=4.286, P = 0.038) and an internal relevant between emotional violence vs. sexual violence (chi2 = 4.239, P = 0.039), physical violence vs. sexual violence (chi2 = 4.482, P = 0.034). The influential factors on the incidence rates of violence would include: sex, smoking status, family without harmony, tendency of bullying others and the level on self-esteem etc.

When the random effect of class difference was separated by multivariate multilevel model, the estimated results would be more precise. Other than paying more attention to both individual and family influential factors when taking measures to reduce the incidence rate of violence in high school students, the effect of environment in the class should not be ignored.