Findings, implications and lessons learned

Various lessons on study design can be learned from this case study.

• Firstly, as has already been mentioned, group feeding means that individuals in a group cannot be assumed to be independent in relation to such measurements as weight gain or feed conversion efficiency. It is possible to be a little more relaxed in the case of mortality and assume that deaths occur independently, provided there is no evidence of transmissible infection within the group. For any variable that is to be measured it is important beforehand to understand what will determine the experimental or observational unit for statistical analysis.
• This study was part of the requirements for a Bachelor of Science in Animal Science degree. It has demonstrated how advisable it is, even in the smallest of degree projects, for a student to seek statistical advice before embarking on his or her study. Other case studies in this Teaching Resource focus on different aspects of study design and a student should look at the various case studies to see the examples that may be relevant to his/her project.

• For this case study a more appropriate experimental design would have been to randomly assign the chicks to a number of smaller surrounds for each feed. Since three levels of spent yeast were to be compared with the control diet and 200 chicks were available, 16 chicks could have been randomly assigned, say to each of 12 surrounds, with four surrounds per diet. This would have used 192 of the chicks and resulted in this skeleton analysis of variance for a completely randomised design.