Effect of Misclassification on Test of Independence Using Different Randomized Response Techniques

Abstract

In Survey Sampling, while we are dealing with sensitive issues, it is a challenge to get the accurate and unbiased answers from the respondents. To fulfill this need researcher exert on discovering such methodology which can help to get accurate responses from respondents in case of sensitive issues. One of the most famous techniques amongst these techniques is randomized response technique. In current research we discuss two dimensional tables and derive the misclassification probabilities for various randomized response techniques. The transition matrices of conditional misclassification probabilities are used to get perturbed or misclassified data and then Chi-square test of independence between two attributes is carried out. The results of chi-square test of independence calculated for perturbed data show that variables are found to be dependent which were independent originally, depicting that misclassification can change the status of dependence in the data. This paper has great contribution towards checking independence status of data while dealing with sensitive issues, where data can be misclassified. The derived matrices of conditional misclassification probabilities can be used acquire estimates of log-linear model for numerous randomized response techniques.