We constructed a measure of general involvement in crime based on the results of a series of twoparameter logistic item response models. These models were estimated separately for each age from 10 through to 18 by pooling responses for all respondents of the same age (irrespective of their years of birth and interview). As a result, the latent score (i.e., theta) reflects the extent to which each individual was involved in crime and delinquency relative to other children and youth of similar age. Item response models offer several advantages over traditional summative crime and delinquency scales based on selfreport items (Osgood, McMorris, & Potenza, 2002). Aside from resolving problems associated with the highly skewed distribution of offending and differences in the seriousness of items, item response models can be used even if respondents have missing data on some items, if some respondents are asked different questions, or if the items change over time. This makes it an especially useful method for analyzing involvement in crime and delinquency in large, longitudinal surveys in which the same items were not used consistently over time—as happened in the NLSY79.
We constructed a measure of general involvement in crime based on the results of a series of twoparameter logistic item response models. These models were estimated separately for each age from 10 through to 18 by pooling responses for all respondents of the same age (irrespective of their years of birth and interview). As a result, the latent score (i.e., theta) reflects the extent to which each individual was involved in crime and delinquency relative to other children and youth of similar age. Item response models offer several advantages over traditional summative crime and delinquency scales based on selfreport items (Osgood, McMorris, & Potenza, 2002). Aside from resolving problems associated with the highly skewed distribution of offending and differences in the seriousness of items, item response models can be used even if respondents have missing data on some items, if some respondents are asked different questions, or if the items change over time. This makes it an especially useful method for analyzing involvement in crime and delinquency in large, longitudinal surveys in which the same items were not used consistently over time—as happened in the NLSY79.
正在翻譯中..
