The same sensitive-behavior survey is administered online to one group and by live telephone interview to another. The telephone group reports much lower prevalence. What should the researcher examine before comparing groups?