Off-campus UNL users: To download campus access dissertations, please use the following link to log into our proxy server with your NU ID and password. When you are done browsing please remember to return to this page and log out.
Non-UNL users: Please talk to your librarian about requesting this dissertation through interlibrary loan.
Using latent structure analysis to examine the distribution of problem behavior: Discrete types vs. propensity explanations of criminality
Abstract
Does a general propensity toward crime exist in everyone to varying degrees or are there distinct types of people more prone to problem behavior than others? The nature of the distribution of criminality is a theoretical and methodological question that has prompted much debate between criminal career and propensity advocates in the field of criminology. The present analysis attempts to advance this discussion by examining two possible distributions. Specifically, I sought to determine whether combining measures of delinquent acts into a continuous latent trait variable yields better predictions of adult problem behavior than combining these same measures into a latent class variable, comprising a set of discrete classes. Using data from two waves of the National Youth Survey, I derived measurement models of criminality using latent structural methods using five equivalent subsets of indicators of adolescent criminality. The latent class analyses consistently identified three classes, conformists, occasional offenders, and serious offenders, across the five subsets, and Birnbaum's (1968) two-parameter logistic model adequately fit four out of five subsets in the latent trait analyses. I then compared the two methods as to their effectiveness in predicting adult behavior measured 10 years later, including employment problems, relationship difficulties, self reported crime, and alcohol and drug use. While the latent trait measure of criminality outperformed the latent class measures for the majority of the adult outcomes, it did not improve significantly upon predictions made by a summative scale. Results of the two continuous measures were roughly equivalent. The discrete class measures outperformed the continuous versions only in isolated cases. Examination of self-report measures of criminality in this study indicates greater support for a continuous, unidimensional distribution over a discrete, multidimensional one. However, integration of both propensity and career criminal ideas can be achieved through the depiction of a continuos, latent propensity of offending as recommended by Osgood and Rowe (1994).
Subject Area
Criminology|Behaviorial sciences
Recommended Citation
McMorris, Barbara Jolene, "Using latent structure analysis to examine the distribution of problem behavior: Discrete types vs. propensity explanations of criminality" (1997). ETD collection for University of Nebraska-Lincoln. AAI9819701.
https://digitalcommons.unl.edu/dissertations/AAI9819701