The quest of this paper is to propose an improved mixed randomized response model for estimating the proportion of sensitive attribute
in the population. Mangat et al. [17] and Singh et al. [28] pointed out the privacy problem with the Moors [18] model. However their models may lose a large portion of data information and require a high cost to obtain confidentiality of the respondents as pointed by Kim and Warde’s [11]. This led Kim and Warde’s [11] to propose a mixed randomized response model using simple random sampling which rectifies the privacy problem. In this paper we have suggested a mixed randomized response model using Mangat and Singh [13] two – stage randomized response technique in place of second randomization device in Kim and Warde mixed randomized response model where they have used exactly the Warner’s randomized response model. It has been shown that the proposed mixed randomized response model is better than Kim and Warde’s [11] mixed randomized response model. Numerical illustration is given in support of the present study. The proposed model has been also extended in stratified sampling.
