| Objective：Sampling survey which related to sensitive subjects is inevitably encountered.Obtain truthful answers to sensitive questions is really a challenge. Sensitive questionsmean that the content of questions themselves can be considered as intrusive and theanswers to questions are socially undesirable or the meanings of questions have threatof disclosure. Potential threat which leads to the decline of accuracy is bothnonresponse and reporting error. A proposed solution is the randomized responsetechnique (RRT) which was frst conceived by Warner in1965. The RRT was designedfor the purpose of decreasing social desirability bias, guaranteeing confidentiality,improving respondent cooperation and procuring reliable estimates. Since Warnerpublished his frst paper on RRT, many researchers have improved and furtherdeveloped this technique. Efforts have been made to propose various forms of RRTduring the last forty years.In the recent research of RRT, much attention has been paid to both dichotomousand quantitative sensitive questions. However, there is less study on the polychotomoussensitive questions. Simple random sampling is the most common method which is usedfor gathering information on sensitive issues. Sampling method is usually limited tosimple random sampling when surveys are on sensitive issues. The participants inresearch on sensitive issues, are always confined to small areas and usually selected bysimple random sampling. What’s more, simple random sampling may be taken intoaccount even in the analysis of complex sampling survey data. The estimation of samplesize is the key part of a sampling survey. However, topics using randomized responsemodel on sensitive issues in sample size determination is not yet available.The selection of six survey methods, which are the combinations of six RRT and two-stage cluster sampling method, were in the present research. Based on the premisethat the estimators of the population parameters for RRT in the two-stage clustersampling survey were given, sample sizes formulae for two-stage cluster samplingsurvey were deduced in this study, so as to minimize the cost of survey implementationfor a specified level of precision and to provide reasonably precise estimates under theconstraint of a fixed budget. These formulae were suitable for complex sample surveyon a large scale. Preliminary investigation on sensitive behaviors among female sexworkers (FSW) in Xichang urban districts showed a meaningful trends in acquiredimmunodeficiency syndrome (AIDS) and provided related statistic value needed forsample size formulae. According to the deduced formulae in this research, requiredsample size at each stage was calculated to estimate sensitive features of FSW inXichang in a field investigation which will be carried out in2015.Method：1. Six survey methods formed form combinations of six RRT (e.g. Warner RRTmodel, Simmons RRT model, Greenberg RRT model, improved RRT model,multiple-choice sensitive question with single response RRT model, multiple-choicesensitive question with indirect response RRT model) and two-stage cluster samplingmethods. Cochran’s classical sampling theory, total probability formula and variancewere applied to deducing statistical formulae that gave estimators of populationproportion and population variance.2. The Cauchy-Schwarz inequality and the Lagrange function were applied todeducing the formulae for the optimum sample size for two-stage cluster so as tominimize the cost for specified sampling errors and to minimize the sampling errorsunder the constraint of a fixed cost.3. We designed RRT randomized devices and crafted a survey questionnaire withsensitive topics. Behavior characteristics were investigated in two-stage clustersampling method of FSW in Xichang urban districts from May to July2011. The urbandistricts were divided into8districts according to the street location. In the first stage,4districts were selected, in the second stage,4venues in average were selected form eachdistrict. Following the statistical formulae, the collected data were preliminarilyanalyzed to estimate related statistics value needed in the sample size formulae. 4. For all of7sensitive questions in two-stage cluster sampling survey, using twoRRT models (e.g. Simmons RRT model, multiple-choice sensitive question with singleresponse RRT model), optimum sample size were obtained when we limited samplingerrors for the minimum of cost and constrained cost on the minimum of sampling errorsrespectively. All these were based on the both related statistics value and sample sizeformulae deduced in this study.Results：1. This study proposed five types of survey methods which combined five RRTmodels(e.g. Warner RRT model, Simmons RRT model, Greenberg RRT model,improved RRT model, multiple-choice sensitive question with indirect response RRTmodel) with two-stage cluster sampling method. For each method, formulae forestimators of population proportion and population variance were given.2. Two types of RRT models combined with two-stage cluster sampling were usedto investigate behavior features of FSW in Xichang urban districts, the preliminarysurvey results showed as follows:The proportion of having spouse or steady sex partner in addition to commercialsex partner was53.33%, with the standard error of3.32%; the proportion of ceasingcommercial sex service when the FSW were diagnosed as STD was75.42%, with thestandard error of3.00%; the proportion of supporting legal prostitution in China was58.75%,with standard error of5.20%;the proportion of condom use, classified as neverused, sometimes used and always used, were10.25%,13.37%,76.38%during lastmonth, with the standard error of4.47%,4.00%,7.68%respectively; the proportion ofcondom breaking during the last commercial sex was26.88%, and the standard errorwas6.00%;the proportion of results on HIV test, including not being tested for HIV,negative and positive, were53.00%,35.00%,5.85%within one year, with the standarderror of5.29%,6.24%,3.00%respectively; the proportion of STD test results, includingfailure to get STD test, negative and positive, were68.00%,19.65%,5.75%, with thestandard error of4.12%,4.47%,2.45%respectively in recent year.3. Our National Natural Science Fund Project will be launched in2015to conducttwo-stage cluster sampling survey dealing with sensitive characteristics of FSW inXichang. Taking preliminary survey data into account, required sample size at each stage were shown. In the first stage, five districts should be selected (n15). Then inthe second stage, an average of six venues should be selected from each chosen district(n26).Conclusion：Statistical formulae about six survey methods for the estimators ofpopulation proportions and corresponding population variance were given in this study.And two RRT models in a two-stage cluster sampling design were successfullyemployed to preliminarily investigate into FSW population which were the high riskpopulation of STD and HIV in Xichang urban districts. Survey methods and statisticalformulae proved to be effective and reliable, and got broad perspectives on application.Formulae for optimum sample size were deduced to provide sample survey design onsensitive topic with scientific methods. Optimum sample size at each stage wascalculated to carry out two-stage cluster sampling formal survey on sensitive featuresamong FSW using two RRT models, which enjoyed a broad prospect of popularization.Preliminary survey results showed that a large scale of FSW argued that prostitutionshould be legalized. Many FSW reported that they did not receive HIV test at thehospital and condom had broken during commercial sex service. The correspondingdepartments should pay more attention to the serious situation and look for suitable wayof settlement. |
PDF Full Text Request