An Empirical Analysis of Various Transformations on Selection Probabilities Using Multi-Character Surveys Incorporating Randomized Response Techniques
by
Jack Allen
Griffith University
Coauthors: Sarjinder Singh University of Saskatchewan

In our current analysis, alternative estimators for estimating population totals in multi-character survey sampling using randomized response techniques have been investigated. We have demonstrated this application in our model when certain variables have poor positive correlation and others have poor negative correlation with selection probabilities. The estimators proposed by Hansen and Hurwitz (1943), Rao (1966) and Sahoo et al. (1994) under scrambled responses are shown as special cases of the proposed estimators. Therefore, a thorough investigation has been tested and executed through simulations, resulting in the best transformation on the selection probabilities in multi-character surveys, when the scrambled responses are collected on sensitive variables. The bias and the relative efficiency of the proposed estimators with respect to one another have been presented in this article.

Key words: multi-character surveys; poorly correlated variables; estimation of total; auxiliary information, sensitive characters.

Date received: July 31, 2001