Question 7 (14 marks) (a) Discuss when it is best to use a completely randomised design, complete block design, or Latin square design. (b) For a complete block design, why do we t an additive model and not an interaction model? (c) Write down a design matrix and parameter vector for a balanced incomplete block design for a model with 3 treatments and 3 blocks, each of size 2. (d) Calculate the reduced design matrix X2|1 for this model. (e) Do you expect the reduced normal equations for this model to have the same solution as the normal equations for a completely randomised design of 6 experimental units over 3 treatments