Abstract: Many researchers investigated adolescent aggressive behavior, in consideration of the negative consequences and psychological problems that aggression will cause. Some researches have been done to show the development of aggressive behavior among adolescence or children, others explored the relationship between aggression and personality, cognitive factors or parenting. With regard to the development of aggression, statistics analysis that deal with continuous variables prevailed researches. Some conclusions was made through regression, ANOVA, HLM or LGM. But researchers early stressed that patterns of aggression should be more important issue and aggressive researchers should explore the transitions of the patterns. We can recognize and intervene the aggressive behavior more comprehensively and adequately if we find the pattern of aggression. Despite several conclusions have been come through analyzing continuous variables, these conclusions are singular and tedious. In more specific terms, the researches adopted continuous variables can indicate the quantitative relationship or the change on the degree or the frequency of aggression. The current aggression research needs some conclusion from the perspective of categorical variables. Such study can show the pattern of aggression behavior. The latent class model and latent transition model that are new analysis method in psychological study, fortunately, can deal with the categorical variate and present the qualitative relationship and change. This study describes patterns of youth aggression and change in aggression profiles. 276 participants attend this longitudinal study. Aggression behaviors and friendship quality were assessed. Through recoding the aggressive scores, we get a set of dichotomous variables that indicated whether specific aggression behaviors is present or not. Latent class analysis was used to determine the aggressive pattern of youth, and latent transition analysis was used to examine the conditional probabilities of transitions in class membership between adjoin time. The results indicate that there are three aggressive patterns in adolescence and the transition of the patterns was rather steady. Three aggression patterns were defined as high aggressive, low aggressive and mixture pattern, respectively. The high aggressive pattern indicated that there are common more frequent behaviors in three aggression forms. On the contrary, there are common less frequent behaviors in three aggression forms. The mixture pattern is a special one where there weren’t consistent frequency in there aggression behavior. That is to say, some aggressive forms presented high frequency and even more frequent behavior than the high aggressive pattern. Meanwhile, other aggressive forms presented low frequency and even less frequent behavior than the low aggressive pattern. The majority of youth remained high aggressive and low aggressive pattern. There are high probability that low aggressive pattern transited to high aggressive pattern, mixture pattern transited to low aggressive pattern or high aggressive pattern. Importantly, there is gender difference in the conditional probabilities of transition of aggressive patterns. The girls maintain mixture and low aggressive pattern more steady than boys. Conflict resolution can predict positively the stability of low aggressive pattern and predict negatively the stability of high aggressive pattern. In future, more factors can influence the class and the transition probability should be added to the study for undrstanding the adolescent aggression in-depth and comprehensively.

Key words: aggression, Latent Class Analysis, Latent Transition Analysis

摘要： 研究采用潜在转变分析探讨了攻击性初中生的类别转变。276名初中参加了为期一年的短期纵向追踪研究，在一年中分两次报告了自己的攻击行为。用潜在转变模型分析了初二到初三时青少年的攻击类别转变，结果表明初中生有三种攻击模式。研究以潜在转变模型进一步探究了这三种攻击模式的变化，结果发现两种模式具有很强的稳定性，不同模式之间也有一定程度的转变。最后，研究探讨了攻击类别转变的影响因素，结果表明性别与友谊质量可以起到显著作用。针对实际意义，文章最后进行了讨论和总结。

关键词: 攻击行为, 潜在类别分析, 潜在转变分析