Abstract Probability histograms of three quadruple difference series are extracted from 24hours GPS observations of two IGS stations with 5Hz sampling rate based on this, it is pointed out that quadruple difference series are not nearly applicable for stochastic modeling due to requirements of high sampling rate, baseline length and probability distribution assumptions on different stations and satellites. Undifferenced combination observation using both geometry-free and ionospheric-free criteria (short for UGIF) is derived from dual-frequency observations, and then a stochastic modeling method for GNSS undifferenced pseudo-range observations is presented based on UGIF. The proposed method can estimate the stochastic model of a single station and satellite pseudo-range observations with high or low sampling rate in real-time or post-processing modes. Finally a case is implemented to test the validity and rationality of the UGIF method using dual-frequency GPS and GLONASS observations collected at CUMT’s single reference station.