|
Abstract:
|
Researchers often study sensitive topics such as sexually transmitted diseases, mental health, or drug and alcohol abuse. When such questions are asked directly, respondents show a tendency of answering them in a socially acceptable fashion rather than answering truthfully, causing bias in estimates. Many randomized response techniques (RRTs) have been developed to encourage honest responses. However, RRTs have their own limitations, such as difficulty in understanding, higher cost or lack of reproducibility. To overcome some of the inadequacies of the RRTs, non-randomized response techniques (NRRTs) have been proposed. In this study we compare some of the main NRRTs, namely Crosswise Model and Triangular Model, to Ardah and Oral (2018)’s two-stage randomized response technique. More specifically, we compare how different cheating proportions in the population effect the variance and mean square errors from these three indirect questioning techniques using an extensive Monte Carlo simulation study. We also provide a real-life example in which we estimate the prevalence of intimate partner violence since the stay-at-home order was issued in Louisiana using the three methods.
|
