Previous PageNext Page

Data and Methods

Data for this report come from surveys gathered from a random sample of 1000 adolescents between the ages of 13 to 17 who were selected from a targeted list of households. The list of eligible households was created by examining drivers license applications and voter registration lists which indicate households with a higher than usual likelihood of containing an adolescent in the target age group. Although respondents are randomly selected, the sampling frame is not, strictly speaking, a random sample. Nevertheless, in previous research this sampling methodology yielded representative samples which are generalizable to the target population (Volberg, 1993; Winters et al., 1995).

Sample characteristics for the current study are listed below in Table 1.2. For most characteristics, the sample is representative. Some caution should be exercised when generalizing the results of this sample to the non-white population. The proportion of this sample which is Anglo matches census estimates almost exactly. However, the study sample underrepresents certain minority groups, and overrepresents the "other" category. For this reason, and because the percentages of various minority groups are rather small, analyses in this report compare Anglos with non-Anglos (including the "other" category) and should be considered as tentative for the non-Anglos.

Table 1.2. Sample Characteristics

(In Percent)

 

Sample Characteristics

(n=997)

Oregon Census

     

Age3

   

14

24.3

25.4

15

26.1

25.2

16

26.0

24.6

17

23.6

24.8

Total

100.0

100.0

     

Race4

   

White

90.1

90.7

Hispanic

1.7

NA

Native American

2.0

2.0

Asian

1.6

2.9

Black

0.2

2.1

Other

3.7

2.3

Total

99.1

100.0

     

Gender

   

Female

46.0

48.5

Male

54.0

51.5

Total

100.0

100.0

     

In order to test the representativeness of the sample, t-tests for proportions were done to determine whether or not the study sample was significantly different by age, gender, and percent white, from the population estimates provided by the Center for Population Research and Census, 1996; no significant differences were found. However, because gambling was significantly different by county, and not all counties were proportionally represented in this survey, data were weighted by county in order to reflect the actual distribution of population by county. Analyses in this report are based on the weighted data. Additionally, because the rates of gambling participation were based on a sample, they should be considered as estimates and are subject to a margin of error of _ 3% (95% confidence level) for the population as a whole. Subgroup analyses are subject to a somewhat higher margin of error due to smaller sample sizes. Estimates of level 2 and level 3 gambling are subject to a sampling error of _ 2%.

Of the original sample of 1000 respondents, three interviews were dropped from the final sample for failing to complete all SOGS items, or for obvious exaggerations of gambling frequency. Thus, the final sample consists of 997 participants. The response rate for the sample was 38%; the refusal rate was 48%.

3 Because Census Bureau estimates collapse ages 12 and 13 into one group, comparisons were based on ages 14-17.

4 Because Census Bureau estimates use a different methodology for calculating the Hispanic population, census estimates can't be compared directly with our sample estimates. Thus, comparisons were calculated excluding the Hispanic category. Total adds up to less than 100% do to refusals.

Previous PageNext Page