Examining and Converging Parallel Measures in Parent－Child Relationship Data

親子關係資料中平行測量的檢測與合併

This study tries to investigate the issue of parallel measures in parent-child relationship data by applying two parts of analysis. Data are drawn from a survey of 173 adolescents (14-17 years old) and their parents (mothers and fathers) in a single school district of a Midwest state, USA. In the survey, parents and adolescents were asked to answer the same questions of parenting behavior independently. Thus, different parenting constructs from three different sources were obtained. The first part of analysis examines the measurement properties of this set of parallel measures in parent-adolescent data. The commonality as well as distinction in measures from fathers, mothers, and adolescents are detected by applying confirmatory factor analysis （ CFA ） together with a multitrait-multimethod （MTMM）approach. The second part discusses the pros and cons of combining parallel measures in a variety of ways. Under different assumptions of measurement properties, a series of competing models that aggregate parallel measures differently are compared. The results show that both the shared and nonshared portions are important components of the parallel measures from fathers, mothers, and adolescents of parenting behavior. Although parent- adolescent discrepancies are usually emphasized, this study shows the shared perceptions of parenting behaviors explain certain amount of child’s behavior outcome. Besides, the analysis finds that although adolescents’ report of parental monitoring is the best predictor of adolescent alcohol use, using multiple informants and collapsing the family score through latent variables significantly improves the explanatory power. The research findings also suggest that studies using single informant data, even those using child reports, may underestimate the influence of parental behaviors on adolescent outcomes.