DETERMINATION OF THE NUMBER OF REPLICATES REQUIRED TO DETECT A SIGNIFICANT DIFFERENCE BETWEEN TWO MEANS IN RABBIT TRAITS

Abstract

The number of replicates required to detect differences of variable size at several levels of significance, has been determined for growth, lactation and digestion traits from the equation: n ≥ s 2 x t 2 x 2/d 2 . Residual standard deviation (s) values have been derived from results obtained with rabbits in our Research Department. An average value of s = 3.93 g/day has been obtained, for instance, in the case of average daily gain, for the average experimental conditions in our laboratory, and from a model including weaning weight as covariate and litter as block effect. This s value increased up to 5.38 when neither covariate nor block was considered in the model. Based on the s values obtained, the number of replicates needed to detect significant differences between two means for different traits were tabulated for several values of d and levels of significance. These values permit to decide the size of experiment according to the size of the differences to be detected. For example in our average experimental conditions, a sample size of 20 rabbits per treatment would allow to detect as significant differences (P = 0.05) of 3.0 g/day of average daily gain, 20 percentage units of mortality and 2 percentage units of dry matter digestibility during the fattening period.