A psychometric investigation of a self -control scale: The reliability and validity of Grasmick et al.'s scale for a sample of incarcerated male offenders

This dissertation investigates the reliability and internal validity of one of the most commonly used measures of self-control, i.e., Grasmick et al.'s 24 item self-control scale. Using a sample of 651 male offenders residing in the Diagnostic and Evaluation center in Lincoln, Nebraska, this dissertation explores the psychometric properties of Grasmick et al.'s scale by answering the following questions. First, is Grasmick et al.'s scale reliable? Second, does it show observed differences across black and white offender groups? Third, is Grasmick et al.'s scale unidmensional? Fourth, is Grasmick et al.'s scale multidimensional? Fifth, can Grasmick et al.'s scale items discriminate among levels of self-control for a sample of incarcerated offenders? Sixth, do respondents' levels of self-control affect survey responses? Finally, are Grasmick et al.'s scale items invariant across black and white offender groups? These questioned are answered using several analytic methods including Cronbach's reliability coefficients, exploratory and confirmatory factor analyses, and a Rasch rating scale model.;Results from this study lead to several interesting conclusions. First, Grasmick et al.'s scale has high reliability for a sample of incarcerated male offenders. Second, racial differences were observed, but these differences were not what would be expected according to self-control theory. Second, confirmatory factor analysis and a Rasch model confirmed that Grasmick et al.'s scale was not measuring one construct, but was shown to measure six correlated dimensions. Third, a Rasch analysis showed that items were able to discriminate among offenders' levels of self-control. Fourth, a Rasch analysis revealed that level of self-control affects responses to survey items: low self-control offenders have unexpectedly lower scores. Finally, several of Gramsick et al.'s scale items exhibited differential item function or bias across black and white offenders. Directions for future research are discussed.