This work aimed to compare conventional regression methods, deviation from the maximum response, reliability index and AMMI (Additive Multiplicative Models Interaction) in order to estimate stability parameters for milk production in Holstein cows. The results obtained were compared with the results from Toler's method. In total 22,560 lactation data were used, up to 305 days obtained through the Holstein Cattle Breeders Association of Minas Gerais State (HCBAMG), Brazil, from 1989 to 1996. Animals were separated into six genetic groups (31/32, GC1, GC2, GC3, GC4 and PO) and evaluated in 14 environments. The conventional regression method presented classification patterns for environmental variations different from the classification given by Toler's method. The latter was considered more appropriate. The AMMI method was not considered to be efficient for the study of phenotypic stability of the Holstein breed groups. The methods of deviation from the maximum response and reliability index presented similar results and complemented information supplied by the method of Toler. The genetic groups GC2 (2nd controlled generation) and PO (pure from origin) presented the largest average productions and alternated with each other as the most stable for the different methods. The genetic groups 31/32 and GC1 presented the lowest average productions and were consistently the most unstable for all the methods studied.
milk; yields; analytical methods
