ABSTRACT
This chapter presents the basic concepts that are required to construct designs to address directly and efciently the aims of a biological experiment. We rst discuss the choice of treatments and materials; treatments should be determined by the aims and the materials should be chosen according to the frame of reference for the experiment (Section 3.1). These two components must then be combined to produce an appropriate design. A good design takes proper consideration of three statistical principles: replication (Section 3.1.1), randomization (Section 3.1.2) and blocking (Section 3.1.3), to reduce bias and maximize the precision of treatment comparisons. We describe the structure of a design with respect to underlying factors using a symbolic form (Section 3.2). Many experiments will use one of the wide and exible family of standard designs, such as the completely randomized design (CRD) (Section 3.3.1), the randomized complete block design (RCBD) (Section 3.3.2), the Latin square (LS) design (Section 3.3.3), the split-plot (SP) design (Section 3.3.4) or the balanced incomplete block design (BIBD) (Section 3.3.5). Once the structure of the design has been determined, a properly randomized layout can be generated with statistical software (Section 3.3.6).
