Abstract: Clock offset measurements of satellite-ground time transfer are usually affected by outliers due to the impact of ionosphere errors, tropospheric errors, and multipath effects. Therefore, in this paper, we propose an autoregressive model based on Bayesian methods for detecting outliers in the clock offset measurements with the classification variables. Furthermore, the model for estimating the magnitude of outliers is given to correct the clock offset measurements, and solves the problem of outlier estimation by transforming it into a simple least square problem. Different schemes based on the real BDS data were designed to evaluate the performance of the new Bayesian method. We applied the new method ito the fast recovery of the clock offset prediction. Test examples illustrate that the Bayesian methods can detect the outliers effectively and estimate the magnitudes of outliers accurately.