Nonparametric statistical analyses consist on nonparametric tests for genotype × environment interaction and selecting the most favorable genotype based on nonparametric stability statistics applied by plant breeders to multi-environment yield trials. This paper presents a comparison and interpretation of these nonparametric tests and stability statistics. Nonparametric statistical analyses present agricultural researchers and specially plant breeders with different challenges and opportunities, so it is best to handle them to their improving. If plant breeder want to select the most stable genotypes for poor environments according to static concept of stability, it is better to use stability statistics S1, S2, S4, S5, NP1 and NP2. In contrast, if breeder want to select for rich environments, it is better to use statistics which reflect dynamic or agronomical concept of stability (S3, S6, NP3 and NP4). Combining yield and stability in one statistic is the major interest for practical applications in plant breeding. In other word, simultaneous selection for both mean yield and stability is important and so the nonparametric stability statistics of rank-sum, KetRank and Top could be useful for this propose. Also, if a plant breeder intends a simultaneous consideration of mean yield and stability by applying the nonparametric statistics, transformation of the original data is not necessary, because the effect of each genotype should not be eliminated from the dataset. Using nonparametric procedures well allows plant breeders to extract more usable information from their data, thereby increasing efficiency and accelerating progress.
