Resumo

OBISPO, Nestor E; ESPINOZA, Yusmary  e  GIL, José Luis. The crossover design: a design for the experimentation with milk cows. Zootecnia Trop. [online]. 2004, vol.22, n.4, pp.384-401. ISSN 0798-7269.

An arrangement in sequences of Latin squares was demonstrated to be statistically efficient to detect the presence of residual effects in the experimentation with milk cows. With this purpose, the milk production of five cows was simulated in the application of three treatments during three periods of 15 days each. A program was generated to estimate the residual effects, and other three to be executed by the Proc GLM of SAS: one for a complete model where all the present sources of variation were included, other to detect at least a period with residual effect, and finally a model similar to a classic Latin Square without taking into account the residual effects. The sequence of Latin squares allowed to detect the residual effects of treatments and the calculation of the least squares means corrected by this effect.

Palavras-chave : Milk cows; crossover design; residual effect; Latin squares; analysis of variance.