Methods utilizing Lg waves for estimating yields of underground nuclear explosions have attracted considerable attention for their accuracy, claims of portability, and suitability for low yield test ban verification. In this study, the method developed by Nuttli is applied to a large dataset of Nevada Test Site explosions recorded on Livermore's Digital Seismic System. The purpose was to check the reproducibility of Nuttli's results and to verify the accuracy of mb(Lg) to estimate yield at high and low yield levels. Portability of the method is examined from the perspective of variability of the magnitude-yield relations as a function of station site and source geology. The results of this study show accuracies of about a factor of 1.6 in yield at the two sigma level for shots below the water table in source regions exhibiting more or less uniform coupling. At Nevada Test Site, there are three such source regions: Pahute Mesa (areas 19 and 20), northern Yucca including areas 2 and 9, and southern Yucca including areas 3, 4, and 7. There is no apparent degradation in accuracies at low yields (1 to 10 kt). Variations in source coupling were observed as a function of shot location and as a function of the medium's gas porosity for shots in tuff. The effect of gas porosity was not observed for shots in alluvium, which have uniformly low coupling compared to shots in tuff. Variations in magnitude-yield relations were observed from station to station, while the four-station network-average magnitudes are more robust and agreed well with Nuttli's magnitude-yield relations. Thus, portability to other test sites should be considered guardedly optimistic, requiring geologic and seismic data related to coupling in each source region and sufficient numbers of stations (four or five) to estimate a robust network magnitude. Taken all together, the findings of this study lend support to this method as a viable means of monitoring a threshold test ban or low yield threshold test ban.