多主体评定青少年外化问题行为的一致性问题

以529名城乡初一学生为被试进行短期追踪研究，考察了青少年自我、同伴和教师三种主体对学校情境中的青少年外化问题行为评定的一致性问题。采用相关分析、多质多法分析、验证性因素分析和检出率分析对三种主体的评定结果进行了分析比较。结果显示，青少年自我、同伴和教师对外化问题行为的评定只存在中低程度的一致性；通过同伴评定测得的四种外化问题行为群之间的相关性高于通过教师评定和自我报告测得的结果，而自我报告测得的四种外化问题行为群之间的相关性最低。不同报告主体对青少年外化问题行为评定的跨时间稳定性存在差异，同伴评定的稳定性最高，其次为教师评定，青少年自我报告的稳定性最低。通过同伴评定测得的结果对多主体评定的解释率高于教师评定和自我报告的结果，自我报告的结果对多主体评定的解释率最低。三种主体所检出的不同形式的外化问题行为的人数比例没有呈现出规律性的差异，只有通过同伴评定检出的不同形式的外化问题行为者的人数比例随着检出标准的提高而减小

Agreement among Different Informants over Ratings of Adolescent Externalizing Behaviors

Research on children and adolescent problem behaviors have traditionally relied on single informants. These normally are self-reports by the child or adolescent participants or evaluations from the participants’ teachers or parents. Such single source data often yield inconsistent and even unreliable findings. Recent years have seen an increasing awareness of the advantage of the multiple-informant approach in data collection. An increasing number of studies have used this multi-informant approach that yield low to moderate agreement among different data sources. However, these multi-informant studies mainly used teacher ratings in addition to adolescents’ self-reports. Existing studies have seldom used peer reports. Because adolescents interact with their peers on a daily basis, not using peer evaluation as a data source will have missed important insight into adolescent behaviors including problem behaviors such as externalizing and misconduct. Most of the existing studies have only examined concurrent agreements among different informants which have been low to moderate. Cross-time stability and agreement across multiple informants remain to be issues for investigation. Studies on problem behaviors have focused mainly on the more severe forms of problem behaviors, such as aggression and antisocial behavior. Little attention has been paid to the less serious forms of problem behaviors such as rule opposition, property vandalism, and status violation. The present paper reports a short term longitudinal study in which adolescent externalizing behaviors were repeatedly rated by the adolescents themselves, the adolescents’ teachers and their classmates. The interval between the two times of assessment was three months. The Clustering Child Behavior Checklist by Frick et al. (1993) was used, which contains four clusters of externalizing behaviors, namely, aggression, opposition, property vandalism, and status violation. The participants were 529 adolescents from 14 classrooms and their head teachers in three junior middle schools of Shandong Province. Multiple analyses were used. These include correlation, multi-traits multi-methods analysis, confirmatory factor analysis, and detectable-rates analysis. Only low to moderate correlations existed among the ratings of different informants (rs range from -0.01 to 0.46), with the correlation between adolescent self-reports and teacher ratings being the lowest (rs range from -0.01 to 0.17) and that between teacher ratings and peer ratings being the highest (rs range from 0.28 to 0.46). Further analysis indicate that the correlations among the four clusters of externalizing behaviors from peer reports were greater than those from adolescent self-reports or teacher ratings. Correlations among the four clusters of externalizing behaviors from adolescent self-reports were the lowest. The cross-time stabilities of the ratings by the three types of informants as indexed by the correlations coefficients between the two waves of assessment also varied significantly across different informants. Compared with self-reports and teacher ratings, peer ratings yielded most stable and consistent results across the two waves of data over a period of three months. Results from confirmatory factor analysis demonstrated that peer ratings had higher coefficients of determination on the constructs of the four behavior clusters than did either teacher ratings or self-reports. There were also differences across the three types of informants in the identification of adolescents with some kind of problem behavior but these differences did not form any clear patterns. Only on peer ratings, the detectable ratios decreased when a stricter detectable criterion was adopted