You are on page 1of 15

Two-way ANOVA

Chapter 14
Factorial
Designs
• Simple one-way designs don’t capture the
complexity of human behavior; our behavior is
the result of many different influences
• The variables can have unique effects or can
combine with other variables to have a combined
effect
• Allow for greater generalizability of results
• Efficient and cost-effective
• Move beyond the one-way ANOVA which has 1
IV, to a two-way design which as 2IVs
• Get information about the main effect of each IV as
well as the interaction effect
• Will be computing multiple F-ratios
• Can be both between-subjects, both within subjects,
or mixed design
• 2 (levels of A) X 2 (levels of B)
• 3 X 2, 2 X 4, ETC.
• Each combination of factor A and factor B creates a
cell (what we are comparing is the means of each
cell)
2X2
A1 A2

B1 A1B1 A2B1

B2 A1B2 A2B2
3X2
A1 A2 A3

B1 A1B1 A2B1 A3B1

B2 A1B2 A2B2 A3B2


A

A1 A2

B1 B2 B1 B2

C1 A1B1C1 A1B2C1 A2B1C1 A2B2C1

C2 A1B1C2 A1B2C2 A2B1C2 A2B2C2


Two-way Between-subjects
ANOVA
• Assumptions:
• The cells contain independent samples
• DV measures of interval or ratio scores are
approximately normally distributed
• The populations all have homogeneous
variance
• 2-way ANOVA – two main effects and an
interaction
Main Effects
• The main effect refers to the effect of that
factor (I.e., the levels) collapsing across
other factors (averaging across those levels)
• For factor A compute the means for each
column, ignoring factor B, which is
represented by the rows
• Essentially perform a one-way ANOVA for
each main effect
Main Effects
• For each main effect (determined by the number of
IVs) you have a null hypothesis and alternative
hypothesis
• Compute Fobt called FA
• If significant then graph the main effect means,
perform post hoc comparisons, and determine the
proportion of variance accounted for by factor A
• Do the same for factor B, collapsing across factor A
• May have different values for k and n for each factor
Interaction
• Two-way interaction effect is the combined effects
of the levels of factor A with the levels of factor B
• Treat each cell in the study as a level of the
interaction and compare the cell means
• Assess the extent to which the cell means differ
AFTER removing those differences between scores
that are due to the main effects of factor A and B
• Thus, differences due to the combination of A and
B, not each separately
Interaction effect
• The relationship between one factor and the DV
change with, or depends on, the level of the other
factor that is present
• The influence of changing one factor is NOT the
same for each level of the other factor
• If the pattern is the same or the relationship is the
same between the scores and one factor for each
level of the other factor there is NOT an
interaction
Hypotheses: for Interaction
• Ho says that differences between scores due to
A at one level of B equal the differences
between scores due to A at the other level of
B
• Compute another separate F-ratio, graph the
interaction, perform post hoc comparisons on
cell means and compute the proportion of
variance accounted for
Eysenck Study
• Level-of processing (5 levels) and age
differences (elderly may not process as
deeply)
• Thus, a 2 X 5 factorial design
• 10 different groups of participants
• Can assess interaction between age and
encoding condition
Computations
• The one-way ANOVA is the basis – a little more
tedious because you have to compute a lot more
• Compute MS total
• MS within (average variability in the cells), still a
reflection of the error variance; used as the
denominator for all three F-ratios
• Three sources of between-groups variance; three
separate MS (factor A, factor B, and the
interaction)
Computations continued
• Compute appropriate SS then divide by
appropriate df to get the MS

You might also like