Based on the five data sources:GPS RO observations, radiosonde observations, satellite Microwave Integrated Retrieval System (MiRS) products, the National Center for Environmental Prediction (NCEP) Global Forecast System (GFS) analysis data, and the European Centre for Medium-Range Weather Forecasts (ECMWF) ERA5 reanalysis data from January 1, 2016 to October 1, 2020, the observation errors of GPS Radio Occultation (RO) data was estimated by using the three-cornered hat method, moreover, the biases and impacts of error correlations between any two of the above datasets on RO observation error estimation were analyzed. Results show that the newly estimate RO observation errors by the MiRS data are greater than those estimated by RO observations, radiosonde observations, GFS analysis data and ERA5 reanalysis data. The RO observation errors decrease toward high latitudes. This study is of great significance to the proper applications of the upcoming RO data.