Appendix 3.4.5: BBSS Sample Size Calculation
Calculating the BBSS sample size requires several inputs. It is often helpful to use a variety of inputs to see the resulting sample sizes. Look to the secondary data review for information on the sample sizes of previous studies in your country, and whether those samples were attained. This background information will help you to tailor the sample size inputs to result in an achievable sample size. Although many research projects set high goals for the sample size, it is best to select a sample that you believe can realistically be attained.
A commonly used sample size formula:
where “N” is the required sample size, “t” is the confidence interval level, “p” is estimated prevalence, “m” is the margin of error and DE is the design effect.
The table below shows required sample sizes for varying design effects. A design effect greater than two is recommended, but the resulting sample size may be prohibitively large for some contexts.
Table1: Sample Size Calculation
Precision / Estimated HIV Prevalence / Confidence Interval / Base Sample Size / Design Effect / Final Sample Size /5% / 30% / 95% / 323 / 1.0 / 323
5% / 30% / 95% / 323 / 2.0 / 646
5% / 30% / 95% / 323 / 3.0 / 969
It is also important to consider the size of the Key Population in your country when calculating the sample size. Some experts suggest that the goal sample size should not be larger than 10% of the size of the target group, especially if the group is considered stigmatized or marginalized, or there is another reason why they would be less likely than the average person to participate in a survey. If you don’t already have population size estimates for the population of interest, estimates are available. For example, UNAIDS estimates that between one and five percent of the male population is likely to be men who have sex with men. It has been estimated that between 0.1 percent and 7.4 percent of a country’s female population are sex workers, depending on the geographic region (Vandepitte 2006).