Protection of privacy with objective prior distribution in randomized response

Abstract

Using objective Bayesian methodology and Hartley entropy, we analyse the method of randomized response with two questions, "intimate" and "alternative non intimate", whose probabilities of response (Yes or No) are "unknown" and "known" respectively. A maximization of conditioned entropy is the criterion for selecting the alternative question to protect the privacy of the interviewed people, for several given "non intimate" questions.