Abstract: Anomaly of mathematical model in Kalman filter can be caused by various outliers in precise point positioning.To find out the minimal detectable biases(MDB) resulting from this anomaly and the effects of undetectable outliers on estimator,the reliability of Kalman filter is necessary.This reliability of both measurements and predicted state are derived according to the reliability theory of the least squares and compared with the reliability of traditional redundancy.The results show that the internal reliability of two methods is related not only with the precision of the observation,but also with satellite geometry as well as elevation.Meanwhile,the internal reliability of predicted state is more sensitive with geometry than the internal reliability of measurements.Although the external reliability of two methods is at mm level,the convergence of pseudorange is faster than carrier phase.