The Global Positioning System (GPS) radio occultation (RO) remote sensing technique has emerged as a promising global observing system. GPSRO data (e.g., phase, amplitude, bending angle, refractivity) are non-traditional meteorological measurements. For an effective assimilation of such observations into a numerical weather prediction system, we need to properly specify the observational errors. The observational errors are affected by many factors, including the observables (e.g., refractivity or bending angles), the observation operators (which relate the model variables to the observables), representativeness error (which are affected by the data assimilation systems' horizontal and vertical resolutions), and observation data quality control and data processing. In this paper, we present the analysis of observational errors with three different observational operators, including refractivity (N), log transformation of refractivity (lnN), and non-local excess phase (EP), that are derived from GPSRO refractivity data for a regional WRF/DART data assimilation system with a horizontal resolution of 36 km. Two expressions of the neutral atmosphere refractivity are tested, one is the traditional two-term equation which relates the refractivity to the pressure, the temperature, and the water vapor pressure; the other one is a newly-introduced expression which links the refractivity to the temperature, and the densities of the dry air and water vapor in the air. One summer month of GPSRO refractivity data from the COSMIC mission over the north western Pacific region is used. The analysis of observational errors from this study and its implication for regional data assimilation will be discussed.