Chain-referral sampling is a sampling method in which sample units are selected based on relationship among population members. It can be applied for sampling rare or hidden population for which sampling frame are not available. However standard statistical methods are capable of producing unbiased estimators for some rare population, but they are not applicable in estimating the features of hidden population like estasy drug users, homosexuals , injection drug users and etc. In standard techniques, it is necessary to select the samples by exact probability. In most of the cases , this obligation means having a sampling frame ,however hidden population lacks this kind of frame. In usual methods of research ,researchers start calculating the features in population from the sample, directly. This leads to production of biased estimator in hidden population . This thesis, first explores chain-referral sampling with probable basis called network sampling which uses standard statistical methods for prevalence rate of some rare qualitative aspects of population .Then, it elaborates on respondent-driven sampling which is a sub-part of chain-referral sampling .This method allows the researcher to investigate and calculate unbiased estimator in hidden population .In respondent-driven sampling , after sampling from population ,this sample is used for gathering some information regarding social networks . This information is then utilized to estimate intended feature in target population. Available software for this method is demonstrated. At last, the necessary cods in S-Plus for simulation study are developed with which the desired properties of respondent-driven sampling are studied.
