Bullying behaviors among US youth: prevalence and association with psychosocial adjustment.

CONTEXT: Although violence among US youth is a current major concern, bullying is infrequently addressed and no national data on the prevalence of bullying are available. OBJECTIVES: To measure the prevalence of bullying behaviors among US youth and to determine the association of bullying and being bullied with indicators of psychosocial adjustment, including problem behavior, school adjustment, social/emotional adjustment, and parenting. DESIGN, SETTING, AND PARTICIPANTS: Analysis of data from a representative sample of 15 686 students in grades 6 through 10 in public and private schools throughout the United States who completed the World Health Organization's Health Behaviour in School-aged Children survey during the spring of 1998. MAIN OUTCOME MEASURE: Self-report of involvement in bullying and being bullied by others. RESULTS: A total of 29.9% of the sample reported moderate or frequent involvement in bullying, as a bully (13.0%), one who was bullied (10.6%), or both (6.3%). Males were more likely than females to be both perpetrators and targets of bullying. The frequency of bullying was higher among 6th- through 8th-grade students than among 9th- and 10th-grade students. Perpetrating and experiencing bullying were associated with poorer psychosocial adjustment (P<.001); however, different patterns of association occurred among bullies, those bullied, and those who both bullied others and were bullied themselves. CONCLUSIONS: The prevalence of bullying among US youth is substantial. Given the concurrent behavioral and emotional difficulties associated with bullying, as well as the potential long-term negative outcomes for these youth, the issue of bullying merits serious attention, both for future research and preventive intervention.

Aitken-based acceleration methods for assessing convergence of multilayer neural networks.

This paper first develops the ideas of Aitken delta(2) method to accelerate the rate of convergence of an error sequence (value of the objective function at each step) obtained by training a neural network with a sigmoidal activation function via the backpropagation algorithm. The Aitken method is exact when the error sequence is exactly geometric. However, theoretical and empirical evidence suggests that the best possible rate of convergence obtainable for such an error sequence is log-geometric. This paper develops a new invariant extended-Aitken acceleration method for accelerating log-geometric sequences. The resulting accelerated sequence enables one to predict the final value of the error function. These predictions can in turn be used to assess the distance between the current and final solution and thereby provides a stopping criterion for a desired accuracy. Each of the techniques described is applicable to a wide range of problems. The invariant extended-Aitken acceleration approach shows improved acceleration as well as outstanding prediction of the final error in the practical problems considered.