A Bayesian Approach to Randomized Response Models by Greta Laesch ’25

Abstract:

In survey sampling, problems often arise when the research pertains to sensitive or stigmatized topics like criminal history, drug use, or sexuality, as respondents may falsify their answers or choose not to respond due to fear or stigma. Traditional techniques like the Warner randomized response model offer a partial solution but have limitations, including unreliable estimates. This colloquium presents a Bayesian approach to address these issues, utilizing flexible priors and robust inference methods. Through an interactive demo and case study, we showcase how Bayesian models can enhance accuracy in estimating sensitive population parameters, offering a practical tool for survey design in challenging research contexts.