A Comparative Study of Modified Hidden Logits Using Randomized Response Techniques

Abstract

The survey sampling is one of the driving and most extensively used technique to collect the data about individual’s behaviors, beliefs, views and opinions on a certain matter or topic. We aim to acquire flawless and reliable responses while collecting data.  This aim is not achieved in such cases, when we are dealing with sensitive or socially stigmatized variables. Frequently respondents give elusive or false or non-responses about sensitive questions. In such sensitive or stigmatized characteristics, we use randomized response techniques (RRT). In current article using Mangat and Singh (1990) randomized response model, a modified hidden logit estimation procedure is presented. The proposed logit estimation procedure is also compared with ordinary logits and Corstange (2004) randomized response model. We detect that modified hidden logit estimates for Mangat and Singh (1990) are closer to the true parametric values as compare to the higher values of p and T and show elevated precision. The akaike and schwarz information criterion are renowned measures to model selection that favors more parsimonious models over more complex models. This study is also conducted for checking best model selection. This paper has a great contribution towards application and estimation of logistic models when sensitive or stigmatized issues are under consideration.