Weather radars provide quantitative precipitation estimates (QPEs) with seamless spatial coverage that can complement limitations of sparse rain gage measurements, including those affecting intensity-duration-frequency (IDF) relations used for infrastructure design. The goal of this M.S. thesis is to assess the ability of 4-km, 1-h QPEs from the Stage IV analysis of the Next-Generation Radar (NEXRAD) network to reproduce the statistics of extreme precipitation (P) in central Arizona, USA, using a dense network of 257 rain gages as reference. The generalized extreme value (GEV) distribution is used to model the frequency of annual P maximum series observed at gages and radar pixels for durations, d, from 1 to 24 h. Estimates of P quantiles from radar QPEs are negatively biased (-20% – -30%) for d = 1 h. The bias tends to 0 and errors are small for d ≥ 6 h, independently of the return period. The presence of scaling for the GEV location and scale parameters, needed to apply IDF scaling models, was found for both radar and gage products. Regional frequency analysis methods combined with bias correction of the GEV shape parameter allow reducing the statistical uncertainty and providing seamless spatial distribution of P quantiles at daily and subdaily durations that address limitations of current IDF relations in southwestern U.S. based on NOAA Atlas 14.