Page 63 -

48 CHAPTER 3 Experimental design

reasonable estimation of the timeline of the experiment and the budget. The basic
structure of an experiment can be determined by answering two questions:

• How many independent variables do we want to investigate in the experiment?
• How many different values does each independent variable have?
The answer to the first question determines whether we need a basic design or a
factorial design. If there is one independent variable, we need only a basic one-level
design. If there are two or more independent variables, factorial design is the way to go.
The answer to the second question determines the number of conditions needed in the
experiment (see Figure 3.2). In a basic design, the number of conditions in the experi-
ment is an important factor when we consider whether to adopt a between-group or
within-group design. In a factorial design, we have a third option: the split-plot design.
Again, the number of conditions is a crucial factor when weighing up the three options.

Design study

Number of independent variables >1?

No Yes

Basic design Factorial design

Number of values in each
independent variable?

Determine number Determine number
of conditions of conditions

Between group Within group Between group Within group Split-plot
FIGURE 3.2
Determining the experiment structure.
In the following sections, we first consider the basic design scenarios involving
one independent variable and focus on the characteristics of between-group design
and within-group design. After that, we consider more complicated designs involving
multiple independent variables, to which understanding split-plot design is the key.

3.3 INVESTIGATING A SINGLE INDEPENDENT VARIABLE
When we study a single independent variable, the design of the experiment is simpler
than cases in which multiple variables are involved. The following hypotheses all
lead to experiments that investigate a single independent variable: