An Analysis of Bullying and Suicide in the United States using a Non-Gaussian Multivariate Spatial Model

Mary Margaret Ryan


Bullying affects thousands of students across the United States (US) each year, which can lead to mental health problems, and in some cases suicide. Intuitively, rates of bullying and suicide may be correlated, and this relationship may change based on region. Thus, a spatial analysis of bullying and suicide rates could help identify regions where more attention is needed to prevent both bullying and suicide. As such, we develop a non-Gaussian multivariate spatial model to analyze bullying and suicide rates in the US. This model incorporates the right-skewed nature of bullying and suicide rates, and leverages multivariate spatial dependence to improve spatial predictions. We apply our statistical model to data obtained from the Centers for Disease Control and Prevention’s (CDC) Youth Risk Behavior Surveillance System (YRBSS). In particular, we consider YRBSS estimates of attempted suicide and (self-reported) bullying rates (per 100 thousand) over the 48 contiguous states. Our model provides accurate spatial predictions for suicide and bullying rates, while providing spatial prediction variances. These results indicate regions that have higher rates of bullying and suicide, which can have implications on policy decisions.


Bayesian; Spatial; Youths

Full Text: PDF


  • There are currently no refbacks.