Here's one that provides discussion fodder...
http://westerncriminology.org/documents ... urigio.pdf
Seems to be a chicken and egg deal - Police tend to disrespect black/hispanic kids more, kids react poorly.
Experiences with the police. A few studies have
found that treatment by the police (respect versus
disrespect) was an important predictor of juveniles' attitude
toward the police (e.g., Friedman et al. 2004). In the
present investigation, students' experiences with the police
were measured by using a set of dummy variables. The
dummy variables differentiated students not stopped by the
police, students stopped and respected by the police, and
students stopped and disrespected by the police. Students
who were stopped and respected by the police were used
as the comparison group for the analyses.
The variables used for not being stopped, being
stopped and respected, or being stopped and disrespected
by the police were generated from questions that followed
a skip pattern in the survey instrument. Students were
asked if the police had ever stopped them. Students were
then asked whether they were respected or not during the
stop. Ignoring the skip pattern in the survey would have
introduced incidental selection bias into the model.
Incidental selection bias is a methodological artifact that
occurs when data are dropped from an analysis in an
artificial (incidental to the method) instead of a random
(non-artifactual) process.
Students who had not been stopped would have been
excluded from the analysis through the incidental selection
process. These students might be different from students
who had been stopped on characteristics related to the
study's outcomes; the not-stopped students would have
been missed in the analysis unless they were captured by
the survey structure and coding of the data. As mentioned
above, dummy variables were created to prevent incidental
selection bias and ensure that the entire sample of students
(stopped and not stopped by the police) was included in
the analyses.
The data were reviewed for inconsistencies in
participants' responses and for coding errors that resulted
from the survey's skip pattern. For example, some
participants responded that they had not been stopped by
the police but then indicated that they had been respected
or disrespected by the police. Other variations in responses
also created inconsistencies. A review of the data
identified 37 cases (4%) with inconsi
