Testing the Effects of Volunteerism and Masking in Respondent-Driven Sampling

Respondent-driven sampling (RDS) has been increasingly applied to sampling illicit drug use populations. RDS is capable of generating a more representative sample of a “hidden” population than the traditional methods, such as snowball and targeted sampling, because of the controls on volunteerism and masking imposed on the recruitment process. In the current practice of RDS sample analysis, sample recruitment patterns are used as the estimates of personal network compositions that are used for population proportions estimation. The precision of replacing network compositions with sample recruitment patterns relies on the assumption of no significant effects of volunteerism and masking. Although this assumption serves as the basis of RDS analysis, testing of this assumption has been challenging. In this study, the authors will present a SAS macro program to conduct such a test in RDS sample analysis, using bootstrap method. Real data will be used for demonstration.