Modifying Classical Randomized Response Techniques with Provision for True Response

Abstract

In this paper we consider four well-known randomized response models; the unrelated question model, the unknown repeated trial model, the forced response model and a model for quantitative responses. Traditionally these models are applied to the respondents chosen by simple random sampling with replacement. We show that a modification may be applied to them and irrespective of how a sample is drawn, we may gain in efficiency in estimation. We illustrate our findings through a numerical exercise.