Study design

3. Appropriate sample size is an important issue to address, especially as large animal to animal variation can occur. Knowledge of likely error variation from previous experiments can help to determine a suitable sample size for each group. In this experiment it was deduced, rightly or wrongly, that eight goats would be sufficient per group. Should there be no significant interactions then the experiment would result in more than adequate sample sizes of 24 goats for comparisons of average genotype differences on the one hand, and of average feeding regime differences on the other. The summing of one factor over another illustrates a powerful feature of factorial experiments. In this case, however, the researcher is interested in determining optimal feeding regimes for each genotype and it is likely that significant interactions could occur. It is important, therefore, to ensure that adequate numbers of goats are included for comparisons between individual groups.

4. When a researcher is discussing an experimental design with a colleague or biometrician he/she should be expected to be asked to justify reasons for his/her decisions not only on sample size but also on questions on numbers of treatments, doses etc. Thus, it would not be unreasonable to discuss the numbers of different genotypes that are proposed for comparison or different levels or types of feeding regimes. For example, the researcher might have wished to compare four genotypes but have realised that experimental constraints prevented this. The option of using two feeding regimes and four genotypes could, however, be explored. In this particular case the 3x3 experiment proved to be the most desirable arrangement, but in other situations adjustments to the initial proposal for the design can be fruitful.