Abstract. We have performed an analysis of reprocessed GPS/MET data spanning 1995–1997 generated by CDAAC in 2007. CDAAC developed modified dual-frequency processing methods for the encrypted data (AS-on) during 1995–1997. We compared the CDAAC data set to the MERRA-2 reanalysis, separately for AS-on and AS-off, focusing on the altitude range 10–30 km. MERRA-2 did not assimilate GPS/MET data in the period 1995–1997. To gain insight into the CDAAC data set, we developed a single-frequency data set for GPS/MET, which is unaffected by the presence of encryption. We find excellent agreement between the more limited single frequency data set and the CDAAC data set: the bias between these two data sets is consistently less than 0.25 % in refractivity, whether or not AS is on. Given the different techniques applied between the CDAAC and JPL data sets, agreement suggests that the CDAAC AS-on processing and the single frequency processing are not biased in an aggregate sense greater than 0.25 % in refractivity, which corresponds approximately to a temperature bias less than 0.5 K. Since the profiles contained in the new single frequency data set are not a subset of the CDAAC profiles, the combination of the CDAAC data set, consisting of 9,579 profiles, and the new single-frequency data set, consisting of 4,729 profiles, yields a total number of 11,531 unique profiles from combining the JPL and CDAAC data sets. All numbers are after quality control has been applied by the respective processing activities.