Abstarct: In most applications including analysis of rate responses and proportions like inflation rate, growth rate, punctual customeres ratio in a financial business and a disease percent in a special region, we are interested in detecting effective elements on response variable. This aim is usually obtainable by regression models. The scale of such responses is continuous and restricted to the interval (0,1). Logistic and probit regression models are the popular models to analyze this type of responses. However, the real distribution of such responses are usually greatly skewed and then the aformentioned models are not appropriate for analyzing them. A new proposed regression model that has a logical and appropriate adaptation with the essence of rate responses is beta regression model which has recently become popular due to high flexibility of beta distribution in modeling various kinds of skewnesses. The beta regression model is an element of generalized linear models. An essential assumption in generalized linear models is independency of response variables. However, in various situations, e.g. in analyzing longitudinal data, spatial data and time series, the response variables are dependent. In these cases, generalized linear mixed models have been used in which the dependency structure of responses is considered by using latent variables. Fitting these models is easier by the Bayesian approach. This simplicity is due to the revolution of MCMC sampling methods. Therefore, our main aim is Bayesian analysis of dependent responses with continuous scale and restricted support.