Things That You Have Done


Most of the items comprising Things That You Have Done were derived from the National Youth Survey. Additional items were written specifically for Fast Track. The following sources provide information on the conceptualization and classification of delinquent behavior based on the NYS items: Elliot, D.S., Ageton, S.S., & Huizinga, D. (1985). Explaining delinquency and drug use. Beverly Hills: Siegel. Elliot, D.S., Huizinga, D., & Menard, S. (1989). Multiple problem youth: Delinquency, substance use, and mental health. New York; Springer-Verlag.

Abstract

Most of the items comprising Things That You Have Done were derived from the National Youth Survey. Additional items were written specifically for Fast Track. The questions mostly address areas such as physical aggression, stealing, vandalism and substance use. The 32 variables in the elementary dataset were coded as frequencies of how often the child reported committing each act in the past year (if they reported not committing the act, the variable was coded as '0').

Sorting the 32 items into eight non-exclusive categories created subscales for this measure. This process was completely conceptual in nature (based on the work of Elliot, 1989. No empirical assessments were used to assist in deriving subscales for the initial analyses of this instrument. The eight categories are as follows: General Delinquency, Crimes Against Persons, Theft, Vandalism, School Delinquency, Organized Delinquency, Alcohol Use, and Drug Use.

Logistic regressions should be used for analyses on the dichotomized-item mean scores or the dichotomized category scores (if generalized linear models using explanatory variables are desired). Because of the nature of the data, researchers may also want to consider analyses that model distributions of count data with high levels of zero response, such as the zero-inflated Poisson (ZIP) distribution. Methods to handle data such as these are not commonplace and advanced analytic methods would need to be employed.

While transformations are often considered with non-normally distributed data, they were ineffective in normalizing these data due to the degree of skew and zero modality. While log transformations succeeded in bringing in extreme values, skewed distributional characteristics were maintained. Although log-transformed scales were used in an initial outcome report for two of the scales, transformations of raw data are not recommended for analyses of these data.

Dataset Names

Raw Dataset Name: CyM
Scored Dataset Name: TYDySCc

Keywords

Physical Aggression, Theft, Runaways, Smoking, Tobacco, Substance Abuse, Gangs, Vandalism, Alcohol, Drugs, Illegal Activity.