The social survey is one of leading mechanism to measure attitudes, behaviors and opinion soft hehumanpopulation. Oftenrespondentsprovideevasiveanswersaboutsensitive questions. A major cause of non-response and evasive answer are Social Desirability Bias (SDB). It is the idiosyncrasy created by interviewee in answering stigmatize questions unfaithfully in the desire of leaving good impression on the interviewer. Randomized Response Technique (RRT) is one of the leading method to circumvent social desirability bias in personal interview surveys. Randomized response techniques are used to collect trustworthy data, to keep the interviewee privacy, and to evade non-response rate when the asked information is sensitive (e.g., drug use, concerning racism, abortion, delinquency, AIDS, or academic cheating). Mostly, in surveysthestandardRRtechniquesareemployedthatrequireabinaryresponsetoasensitive question, to estimate the proportion of people bearing sensitive characteristics. Furthermore, RRT models allow respondents to mask their actual response by giving a scrambled response which offering them complete privacy, as the researcher is able to unscramble responses at an aggregate level but not at an individual level. In this thesis, we developed two-stage, optional and scrambled RRT models under simple random sampling. The utility of proposed RRT models are also explored in stratification, two-stage cluster sampling and stratified two-stage cluster sampling . Efficiency comparisons of proposed estimators are obtained to observe the performance of the estimators. In Chapter 2, we proposed two-stage RRT models to estimate the proportion of stigmatized attribute. In Chapter 3, we proposed optional RRT model to estimate the prevalence of stigmatized attributeandsensitivitylevel. InChapter 4, wepresent two-stageRRTmodels toestimate the mean number of persons possessing a rare sensitive attribute using the Poisson distribution. In Chapter 5, a new scrambled randomized response (SRR) model has been proposed for estimating the population mean of a sensitive variable in presence of scrambled response. In Chapter 6, new partial SRR model has been proposed for estimating the population mean of two quantitative sensitive variables simultaneously. In Chapter 7, we developed Bayes estimators of optional unrelated question RRT model.