The main purpose of this paper is to establish a new respondent driven sampling estimator. The previous estimators predicted the proportion of the interesting group appropriately only under strong assumptions such as small sample fraction and no seed bias. However, in reality, these assumptions do not always hold. Therefore we need to develop a new estimator which is not so sensitive to those assumptions. My study starts from the sketchy idea of Ian Fellow. By revising his idea, I built up a new estimator and tested it under different conditions. By the simulation, I could verify that the new estimator predicts the true proportion of the interesting group better than previous estimators.