Endnotes
1. A large number of respondents (n=882) who reported gang presence inadvertently were not asked the followup question regarding how they knew gangs existed. This omission reduced the number of respondents who should have been asked the followup question from 3,486 to 2,604; therefore, the responses to this question are not nationally representative. The analyses directly affected by this omission pertain to student indicators of gangs (tables 1, 2, and 4). The skipped students were significantly (but not greatly) more likely to be younger, white, from higher income households, in private schools, in suburban areas, and in areas where crime was less prevalent. Thus, the responses of the students who were asked the followup question would skew the gang perceptions of students toward greater seriousness. For this reason, weighted data were not used for the analyses.2. In this study, findings were reported as significant if the Pearson chi-square had a probability of less than 0.001 of occurring by chance or because the estimates were based on a sample. The probability of this value of chi-square occurring as a result of sampling error is less than or equal to 0.001. The chi-square statistic measures the probability that two variables will appear to be correlated simply because the estimate is based upon a sample rather than the population. Usually, researchers are willing to assume that the observed results are not due to sampling error if the probability that the results are due to sampling error is 5 percent or less. The more stringent 0.001 criterion was used to take account of the fact that the NCVS is not a simple random sample and that the standard errors computed in standard statistical packages such as SPSS will be too small for use with the NCVS. The customary response to this problem is to use complex standard error estimation routines (e.g., WESTVAR), or to double the standard errors computed by statistical packages to take account of the cluster in the NCVS sample. Using the more stringent significance level effectively accomplishes the same end.
3. A drug availability index was created for this analysis. Each of the nine substances that was reported to be readily available added "1" to the score on the index.
4. A school security index was created for this analysis. Each of the six school security measures added "1" to the score on the index.
5. Correlation coefficients indicate the strength and direction of the relationship between two variables. More specifically, they indicate the unit change in one variable that can be expected when the value of the other variable changes one unit.
| Previous | Contents | Next |
 |
|
| Youth Gangs in Schools | |