This paper shows, based on daily records, the modeling of maximum precipitations
in each quarter of eighteen meteorological stations located in different parts of
Uruguay. We compared the performance of the classic likelihood ratio test with
one of the truncated Crámer-von Mises type. Most of the stations did adjust under
the Gumbel distribution with few Fréchet and Weibull cases, obtaining a most
appropriate truncated Crámer-von Mises test performance. From the adjustment in
each of the stations and the combination of three statistical techniques (k-means,
Kolomgorov–Smirnov test of equality of distributions and test of independence)
we concluded that the maximum rainfall throughout the Uruguayan territory is
homogeneous with a slight difference between the southern and northern regions.