The objective of this work was to estimate minimum sample size (n) for comparison of treatments in experiments of consumption and digestibility of cattle in a multitrait analysis. Data from digestibility of 72 heifers aging approximately 18 months and with 250 kg of live weight were used. A completely randomized experiment was conducted at Embrapa-Centro de Pesquisa de Pecuária do Sudeste, São Carlos, SP, Brazil, from 1988 to 1989, considering the combination of nine treatments organized in a factorial 3 x 3 (three genetic groups: Canchim, Nellore and ½ Canchim + ½ Nellore, and three dietary levels of crude protein: 6, 10 and 13%, and eight heifers each). The end points analyzed were feed consumption (g/kg0.75), digestible energy, retained nitrogen (RN), RN (mg/kg0.75), and digestibilities of dry matter, crude protein, neutral detergent fiber and acid detergent fiber. The minimum sample size required to detect significant differences (delta) between mean vectors was calculated by a SAS program considering a t-variate normal distribution with mean vector 0 and covariance matrix sigma, Hotelling's T² test, F-distribution with noncentrality parameter (<FONT FACE=Symbol>d²D)</FONT> and combinations of type I error (alpha) and power of the test (1 beta) and delta. The calculated n ranged from 6 to 47, and was affected more by variation in delta than variation in alpha or power of the test.
F-distribution; minimum significative difference; Hotelling's
