Varying police discretion has been assessed in this research by using multivariate methods. Datasets of this research has been obtained from three resources. The first set of statistical and interview data is gathered from over 300 interviews of police officers from more than 100 police agencies across Canada, taking into account all provinces and territories, different communities and police services. Police agencies are selected in stratified random sampling so that proper representation and population size can be ensured. Purposive sampling is used to select interviewees who have worked with the offenders so that representation of all the different police departments and levels can be confirmed. Police officers with different experiences are ensured to have been involved with youth offences and all their interviews are tape recorded. The interview data is quantitatively coded to the level of the respondent, thus dealing with the question of reliability of a tape recorded interview. These selective observations are met by a confirmation bias as officers are asked to compare the aspects of crime and their extent of affecting police discretion with young offenders. In order to minimize errors, they are asked to categorize the factors affecting discretion into a “major factor”, a “secondary factor”, and a “minor factor” or “not a factor”.
Ride-along data is collected from thirty police agencies with young offenders. The duration of time for ride-along data collection ranged from four to twelve hours (Carrington & Schelenberg, 2004). A direct observer is observed and recorded in detailed field notes during and after the ride-along. During ride-along, three types of information are recorded, the details of the incident and its descriptions, police behaviour and discretion. The interviews as well as the ride along data thus serve to eliminate the inaccuracies such as overgeneralization and inaccurate observation.