The objective of this work was to evaluate a method to estimate the number of individuals (n) to be used in experiments involving multivariate analyses of repeated measurements over time on the same experimental unit. The method was tested with data from milk production taken monthly up to complete 10 controls (t = 1, 2, ... , 10) or evaluation conditions of Holstein cows. The estimates of n were obtained by a SAS program considering a t-variate normal distribution, zero mean vector, and covariance matrix sigma, Hotelling T² statistical, F-distribution with noncentrality parameter delta²delta. The link of the observed data with the method is made by means of the variance-covariance matrix. The method estimates the value of n required to detect significative differences (delta) between any two means on each t > 2 evaluation condition, considering combinations of levels of type I error (alpha), power of the test F (1-beta) and delta. For the 10 evaluation conditions the n ranged from 11 to 89 and was affected mostly by variation in delta, followed by alpha and beta.
F distribution; lactation controls; power of test (1-beta); simulation; type I error (alpha)
