The OCJS panel sample of regular drinkers aged 12-29 (N = 2649) was first examined thoroughly, and results for key variables are summarized here and in Table 1 below.
The logistic regression models presented here examine to what extent current and prior binge-drinking frequency are associated with the odds of committing an assault offense in the same year, using data from the OCJS panel respondents (aged 16-29).
However, the complex weighting strategy employed in the OCJS addresses the issues of demographic characteristics and attrition (see Phelps et al., 2007 for further details on the weighting strategy).
However, the designers of the OCJS have been rigorous in adopting methods, such as computer-assisted self interviewing (audio-CASI), which reduce the potential for such bias to occur.
From the OCJS data, arrest is rare, even for serious offenders, and is better predicted by drug use than by the (dichotomous) frequency of offending.
In contrast to the results reported above from the OCJS, the reported frequency of offending was found to be a significant predictor of arrest in the ESYTC (perhaps due to the use of a continuous, rather than dichotomous measure of frequency), but drug use also had an independent effect in increasing the probability of having adversarial contact with the police (with an odds ratio of 2) (McAra & McVie, 2005, p.
The analysis of OCJS data that is presented in table 3 also supports this disciplinary model.