RRTCS: An R Package for Randomized Response Techniques in Complex Surveys

Abstract

Randomized response (RR) techniques may be used to compile more reliable data, to protect the respondent's confidentiality, and to avoid an unacceptable rate of nonresponse when the information requested is sensitive (e.g., concerning racism, drug use, abortion, delinquency, AIDS, or academic cheating). Standard RR methods are used primarily in surveys that require a binary response to a sensitive question, and seek to estimate the proportion of people presenting a given (sensitive) characteristic. Nevertheless, some studies have addressed situations in which the response to a sensitive question results in a quantitative variable. RR methods are usually developed assuming that the sample is obtained using simple random sampling. However, in practice, most surveys are complex and involve stratification, clustering, and an unequal probability of selection of the sample. Data from complex survey designs require special consideration with regard to the estimation of finite population parameters and to the corresponding variance estimation procedures, due to the reality of significant departures from the simple random sampling assumption. In such a complex survey design, unbiased variance estimation is not easy to calculate, because of clustering and the involvement of (generally complex) second-order inclusion probabilities. In view of these considerations, a new computer program has been developed to provide a method for estimating the parameters of sensitive characteristics under a variety of complex sampling designs.