Abstarct: For many applied situations‎, ‎in a regression model‎, ‎the response is a binary variable‎. ‎The general models for analysis of such responses are logistic‎, ‎probit and Cloglog‎. ‎Logistic and probit symmetric lixnk models are not appropriate choices when the number of observations in the two response categories is unbalanced‎. ‎Further‎, ‎the positively skewed Cloglog lixnk model will lose its efficiency when responses are negatively skewed‎. ‎The generalized t-lixnk model is a good alternative in such situations‎, ‎though the constraint on the shape parameter greatly reduces the possible range of skewness provided by this model‎. ‎Wang and Dey (2010) proposed a new extremely flexible model baxsed on generalized extreme value distribution‎. ‎In this thesis‎, ‎we reintroduce generalized extreme value lixnk model and discuss its fitting from Bayesian frxamewrok‎. ‎Due to the complexity of posterior distribution of the model we use Markov chain Monte Carlo methods‎. ‎We explore the efficiency of generalized extreme value lixnk model by a simulation study and demonstrate its applicability in the analysis of a real example‎.