This paper presents a modification of Kim and Warde’s (2004) Stratified Warner’s Randomized Response Technique (SWRRT) and Kim and Elam’s (2007) Stratified Unrelated Question Randomized Response Technique with known distribution (SURRT) using optimal allocation. Our models further reduce the non-response bias by introducing the concept of sub-samples of non-respondent developed by Hansen and Hurwitz (1946) to the above mentioned models. In this paper we perform an empirical practice of our model and we also perform the empirical comparison of our models with both Kim and Warde (2004) and Kim and Elam (2007) models. We discovered a new note in the empirical comparative study of our models.