Published online by Cambridge University Press: 02 September 2010
In comparative digestibility trials with growing pigs Latin squares are widely used as experimental designs. In this paper conditions for the application of Latin squares are explicitly stated and checked with real data. Data were collected at the Institute for Livestock Feeding and Nutrition Research in the Netherlands and consisted of digestibility trials run on pigs of 40 to 100 kg live weight. There is strong evidence for a certain form of period × animal interaction in the data. This may seriously affect estimates of treatment effects and results of significance tests. It is concluded that Latin squares are unsuitable for comparative digestibility trials with pigs growing from 40 to 100 kg. When there is no interest in statistical comparison of the diets under study, Latin squares can still be used. Some suggestions for alternative designs are given.