Abstract:
|
In this paper, we propose a new weighted, squared distance while minimizing the distance between the true proportions and the observed proportions of (Yes, Yes), (Yes, No), (No, Yes), and (No, No) answers in the setup of Odumade and Singh (2009) model.
The resultant estimator is shown to be an unbiased estimator of the proportion of the sensitive attribute of interest in a population and has a smaller variance than the estimator of Odumade and Singh (2009) with the same protection of the respondents.
|