A tutorial on the use of informative Bayesian priors for minority research

Aims. Quantifying differences between minority and majority groups, such as sexual minorities (SM) and heterosexuals, is difficult due to small sample sizes. Bayesian analyses is one solution to addressing small sample sizes in minority group research, whereby previous research can be used to inform our models. In the present tutorial, we offered an overview of Bayesian statistics and described an approach to constructing informed priors using a large survey when estimating values in a smaller survey. In an applied example, we determined whether SMs in New Zealand reported more stress relative to heterosexuals and whether stress mediates the link between SM status and alcohol use.Design. Two cross-sectional, stratified, and nationally representative health surveys from the US (National Survey of Drug Use and Health (NSDUH)) and New Zealand (New Zealand Health Survey (NZHS)).Settings. United States, New ZealandParticipants. We used data from 83,661 (SMs = 5593) survey respondents in the US and 24,098 respondents in NZ (SMs = 619).Measurements. Demographic items (sex, age, ethnicity, sexual identity), the Kessler psychological distress scale, and the Alcohol Use Disorder Identification Test (AUDIT).Findings. Using a larger survey to inform priors reduced the uncertainty of estimates derived from small subgroups in a smaller survey relative to uninformed priors.Conclusion. Informed Bayesian analyses are an important tool for researchers studying minority groups and the application of informative priors allows for more reliable estimates of health disparities.