Handling Unreasonable Data in Negative Surveys

Negative survey is a method of collecting sensitive data. Compared with traditional surveys, negative survey can effectively protect the privacy of participants. Data collector usually has some background knowledge about the survey, and background knowledge could be effectively used for estimating aggregated results from the collected data. Traditional methods for estimating aggregated results would get some unreasonable data, such as negative values, and some values inconsistent with the background knowledge. Handling these unreasonable data could improve the accuracy of the estimated aggregated results. In this paper, we propose a method for handling values that are inconsistent with the background knowledge and negative values. The simulation results show that, compared with NStoPS, NStoPS-I and NStoPS-BK, more accurate aggregated results could be estimated by the proposed method.