Documente Academic
Documente Profesional
Documente Cultură
When an experimenter wants to examine two or more independent variables, he/she must use the
factorial design. In using the factorial design, the experimenter yields more information in less
time (McBurney, 1990). Below are the other purposes of using the factorial design (Simon,
1978; McGuigan, 1979).
How does an experimenter draft a factorial design? (Meyers, 1987; Trohim, 2006)
If the number of levels of each factor is multiplied, the result is the number of treatment groups. As
time in instruction has two levels and class setting also has two levels, there are four treatment
groups in the sample experiment: (1) one-hour in-class session (2) one-hour pull-out class (3) four-
hour in-class session, and (4) four-hour pull-out class.
As soon as the factors, levels, and the number of treatment groups have been decided, the
experimenter must create a table. The values of the dependent variable (data found inside the cells)
should be group averages. Figures 2, 3, and 4 (page 3) show complete contingency tables where the
values inside the cells are the average quiz grades (ranging from 1 to 10) of the students in each
group.
Apart from creating a table, an experimenter can also depict a factorial design by notation. In
notations, the number of numbers tells the number of factors while the value of each number tells
how many levels each factor has. The design shown in Figure 1 is called a 2 x 2 factorial design
because it has two factors with two levels each. Any design with two factors, regardless of the
number of levels, is called a two-factor design. The order of numbers makes no difference; a 2 x
3 factorial design can also be termed a 3 x 2 factorial design.
A main effect is the action of one independent variable in the experiment; one main effect
corresponds to the influence of only one independent variable across treatment conditions
(Meyers, 1987). A main effect is calculated by getting the average of the values across rows
and columns.
What is an interaction?
An interaction occurs when the effect of one factor depends on the level of other factor/s
(Meyers, 1987; Trochim, 2006). Finding interactions is needed to interpret the main effects.
Interactions reveal if the factors produced their main effects independently or with the influence
of another factor. A case in point is the role of moderator variables in an experiment. Moderator
variables interact with the independent variable as they strengthen, weaken, or reverse its effect.
Another case in point is a drug interaction where two drugs are harmful when taken together but
safe when taken separately. Trochim (2006) lists three ways in which an experimenter can
determine whether there is an interaction between independent variables or not.
Variations in a factorial design can be plotted in either a line graph or a bar graph, both to
determine if there is an interaction between or among factors. In making both graphs, the Y-axis
should always be the dependent variable. The X-axis would depend on the factor whose main
effect you want to observe; if you want to observe the main effect of Factor A, make Factor B
the X-axis and vice versa. There should be one bar or line graph per factor to determine whether
each factor has significant main effects. Between a line graph and a bar graph, it is the former
that is commonly used. The examples below come with line graphs.
Figures 3 and 4 show outcomes where only one factor has a main effect and no interaction is
found between the factors. In Figure 3, the lines in the graph for setting overlap each other,
indicating that the effect is null, unlike the lines in the graph for time where they are parallel
indicating that time renders a significant main effect and has no interaction with setting. The
reverse outcome is shown in Figure 4.
Figure 3. A 2x2 factorial design which shows one main Figure 4. A 2x2 factorial design which shows one main
effect [for time] and no interaction. (Trochim, 2006) effect [for setting] and no interaction. (Trochim, 2006)
Figure 5 shows an outcome where both factors have main effects and no interaction is found
between the factors. Because both main effects are significant, it is easier to interpret results. It is
shown in Figure 5 that the best variation is clearly the four-hour in-class session because (1) it
garnered the highest average in the table (2) the blue line, which corresponds to 4 hours and in-
class in each graph, is higher than the red line, which in turn corresponds to 1 hour and pull-out.
It can also be seen in Figure 5 that the lines in both graphs are parallel, showing that the factors
do not interact with each other.
Figure 6 shows an outcome where both factors have main effects and an interaction is found
between the factors. As shown in Figure 6, it can be interpreted that the four-hour in-class
session is superior to the rest for the same reasons cited in Figure 5. It can also be seen in Figure
6 that the lines in both graphs are tangential, showing that time and setting in this case interact
with each other.
Figure 5. A 2x2 factorial design with two main effects [for Figure 6. A 2x2 factorial design with two main effects [for
both time and setting] and no interaction. (Trochim, 2006) both time and setting] and an interaction. (Trochim, 2006)
References: