Title: Privacy protection in surveys with RR techniques using nonparametric regression
Authors: Ismael Sanchez-Borrego - University of Granada (Spain) [presenting]
Maria del Mar Rueda - Universidad de Granada (Spain)
Abstract: People do not often respond truthfully when asked personal or sensitive questions in a survey, like those involving stigmatizing characteristics like regular gambling, marijuana consumption, tax evasion, etc. We consider the problem of estimating the finite population total of a quantitative variable using the randomized response (RR) technique, that preserves the privacy of the respondents. We propose a model-assisted estimator based on nonparametric regression, which can handle discrete and continuous data and is valid for any sampling design. The proposed method is shown to share some theoretical properties with the mixed-data kernel-based smoother in the survey context. We have investigated the practical performance of the proposed method under different scenarios with some randomization devices and different sampling designs. The nonparametric estimator is effective in estimating the population total of a continuous variable in simulation experiments in both natural and artificial populations.