For school surveys, for example, the most selleck screening library cost-efficient sample allocation among stages comes from having data on components of sampling error associated with sampling schools and students within schools, on the average cost of adding another school to the sample, and on the average cost of including another student respondent to the overall sample of students (Cochran, 1977). These items may be extracted from earlier survey waves or from a survey conducted in a comparable country setting. Similar data needs exist for adult surveys (e.g., considering sampling units, households, and the number of respondents in each household). 5. There is also a need to consider sentinel surveillance, given the costs and other challenges involved in obtaining representative sampling in many places.

These designs can also facilitate more rapid response to emerging information needs. This is especially true when the primary objective is to evaluate policies and not to obtain representative estimates of the prevalence of tobacco use behaviors. Although comparisons of sex- and age-specific prevalence estimates obtained using modified sampling strategies with the GATS standard sampling strategy could be informative. 6. Differences in estimates across surveys. There is a need to assess the nature and extent of possible differences in estimates of various tobacco use behaviors and factors influencing use across surveys (e.g., Fidler et al., 2011; Giovino et al., 2012; Hammond, 2009; Jha, Ranson, Nguyen, & Yach, 2002; Pampel, 2008).

Comparisons should take into account multiple factors that influence population-based estimates, such as definition of a user, sample frame, type of survey, and editing procedures (IARC, 2008). Sex- and age-specific analyses of key indicators, especially daily and nondaily use, should be compared. 7. Differences in data collection strategies. As efficiencies in data collection strategies evolve, studies of possible differences in estimates provided by face-to-face, telephone, and online data collection strategies are needed. This need is more relevant in high-income countries, where telephone and Internet coverage would be sufficiently high to conduct such work. A split-sample technique (e.g., CDC, 1994) could compare sex- and age-specific estimates of key tobacco use indicators in random samples. 8. Differences in obtaining cooperation.

In some countries, respondents may feel compelled to participate when asked. In others, participation rates will be substantially lower as people feel more comfortable refusing participation. Among the 14 GATS Wave I countries, response rates ranged from 65.1% in Poland and 76.1% in Ukraine to 97.2% in Egypt and 97.7% in Russia (Giovino et al., 2012). When people AV-951 feel compelled to participate, do they feel more compelled to provide socially acceptable answers? How might differing participation rates influence prevalence estimates? 9.