Once the research question was framed, three phases of research were designed and implemented. In the data collection phase, we employed three different random sampling techniques. First, student researchers and volunteers were deployed with surveys among student-attendees hanging out and waiting in lines at both summer conferences, resulting in 1278 surveys collected in Los Angeles and 2235 surveys collected in Washington, D.C. These surveys were all collected prior to the formal beginning of the conferences, allowing us to establish a baseline profile of student-attendees drawn from this randomly administered sample of more than 10% of participants.

Next, we mailed a survey to a random sample of 1500 names drawn from among all of the DC/LA '97 registrants who will graduate from high school in 1998, 1999, or 2000. These surveys, mailed out in November 1997, brought a return of 126 from LA attendees and 286 from DC attendees. This return rate of 27% fell short of the desired goal of 33%, but was determined to be acceptable because of the later sample to be collected.

Finally, 27 Huntington College students enrolled in Dr. Rahn's Evangelism class conducted phone surveys with 170 LA attendees and 330 DC attendees randomly selected from among registrants who will graduate in 1998, 1999, or 2000.

The strength of this research is based largely on the fact that random sampling procedures were successfully employed through three different techniques resulting in sample sizes large enough to detect differences between groups of persons. These techniques allow us to generalize the findings of the project to all of those student-attendees of the conference currently in high school.

In the data analysis phase of the research, a number of different steps were taken with increasing degrees of sophistication, all intended to help detect differences among the students surveyed. At the outset, frequency distributions for all responses were collected and reported as percentages and histograms. Subsequently, a correlation matrix was developed for all items. The data from each of these approaches were used to help identify which items could be controlled for in the next step—analyses of variance statistical tests. For example, the item “hang out with your non-Christian friends” stood out in the correlations because it had virtually NO correlation with any other item. This led to the decision to include it in ANOVA testing. Separate one-way ANOVAs were employed to determine if evangelism behavior differences exist between students who answered a particular way on the survey (i.e., male vs. female; south vs. midwest vs. west coast; weekly church attendees vs. monthly church attendees vs. seldom church attendees, etc.). A total of 22 different items were controlled for, resulting in a total of 1017 separate ANOVA tests being performed. For each of the tests where main effects were found (indicating the presence of differences), subsequent Scheffé post-hoc analyses were done in order to determine exactly where the differences were located. The significance level of these post-hoc tests was set at .05, meaning that we have a 95% degree of statistical confidence when we speak of existing differences.

These test statistics, while laborious, were chosen because they are both robust and conservative. A robust test means it has a great deal of utility and can be counted on in a variety of settings. The fact that the tests were conservative means that the tests were designed to error by reporting too few differences rather than too many. As was the case with the random sampling procedure, these decisions allow us to have greater degrees of confidence in the findings.

The final phase of the research has been data results and interpretation. This stage is represented by the current report.