Chapter 9 – Factorial Designs Factorial Design -- definition – Two or more IVs – every level of one IV • combined with every level of other IV – IVs -- called factors
Jan 14, 2016
Chapter 9 – Factorial Designs
Factorial Design -- definition
– Two or more IVs
– every level of one IV
• combined with every level of other IV
– IVs -- called factors
Example: memory for words
List A
key
tree
door
hair
wood
pan
sink
wall
dirt
ball
List B
computer
microwave
elevator
telephone
stereo
spatula
camera
cabinet
waterfall
basketball
List C
walk
jump
read
drink
eat
touch
talk
sit
write
dance
List D
remember
liberate
jeopordize
violate
sanitize
stimulate
dominate
prosecute
hypnotize
calculate
Factors
word type (noun or verb) word length (short or long)
Numerical notation
Denotes:
– Number of IVs (factors)
– Number of levels of each factor
e.g. 2 x 2 design e.g., 4 x 3 design
Factorial Table
One factor rows – Other factor columns Each column / row corresponds to a level of its factor Each cell represents a specific condition Combination of specific levels of each factor note: #conditions = levels x levels
Word Length
Short Long
Word
Type
Noun
Verb
Alternative labeling scheme
word type = factor A length = factor B)
4 Conditions
– noun / short: A1B1– noun / long: A1B2– verb / short: A2B1– verb / long: A2B2
Results of Factorials: Main effects & Interactions Main Effect
– effect of one IV independent of other IV– collapse across levels of other IV– compute and compare marginal means
4.756.75
5
6.5Word
Type
Word Length
46Verb
5.57.5Noun
LongShort
Results of Factorials Interaction between factors (IVs)
– Ask: does effect of on IV depend on the level of the other IV?
– If the answer is “yes” – you have an interaction– easiest way to see interaction -- in a line graph
0
1
2
3
4
5
6
7
8
Short Long
NounVerb
Results of Factorials Can plot either way -- same interpretation
0
1
2
3
4
5
6
7
8
Noun Verb
ShortLong
Example 2: Effect of violent media on 6th grade children
IV1: Comic book content (violent vs. nonviolent) IV2: Sex of subject (male vs. female) DV: Measure of aggressiveness (response to scenario)
3.256.5
6
2.75Sex
Comic Book
48Boy
2.53Girl
NonViolentViolent
Interaction
Does type of comic book have effect? -- it depends Does sex matter? -- it depends
0
1
2
3
4
5
6
7
8
9
Violent Nonviolent
GirlBoy
Interaction
Can plot either way
0
1
2
3
4
5
6
7
8
9
Girl Boy
ViolentNonviolent
Possible patterns
can get any possible combination of main effects and interactions
main effects but no interactions interactions but no main effects main effect of one factor, but not the other
**INTERPRETATION:
– INTERACTION SUPERCEDES MAIN EFFECTS
0
1
2
3
4
5
6
7
8
9
Violent Nonviolent
GirlBoy
5.55.5
8
3Sex
Comic Book
88Boy
33Girl
NonViolentViolent
0
1
2
3
4
5
6
7
8
9
Violent Nonviolent
GirlBoy
38
5.5
5.5Sex
Comic Book
38Boy
38Girl
NonViolentViolent
0
1
2
3
4
5
6
7
8
9
Violent Nonviolent
GirlBoy
33
3
3Sex
Comic Book
33Boy
33Girl
NonViolentViolent
0
1
2
3
4
5
6
7
8
9
Violent Nonviolent
GirlBoy
55
7
3Sex
Comic Book
68Boy
42Girl
NonViolentViolent
0
1
2
3
4
5
6
7
8
9
Violent Nonviolent
GirlBoy
66
6
6Sex
Comic Book
48Boy
84Girl
NonViolentViolent
40
50
60
70
80
90
100
Low Medium High
EasyHard
Anxiety Level
Low Medium High
Test Difficulty
Easy 90 85 90
Hard 80 75 50
Types of Factorials
Between Subject Factorials Within Subject (Repeated Measures) Factorials Mixed Factorials IV x PV Factorials
Between Subjects Factorial
all factors(IV) are manipulated between subjects e.g., study material by material generation (2x2 between
subject design)
– study material – outline vs. questions
– material generation – self vs. instructor
– four conditions: four different groups of subjects
concerns: begin with equivalent groups
– random assignment
– matching
Within Subjects Factorial
all factors(IV) are manipulated within subject
e.g., effect of sex and age of approacher on personal space (2x3 within sub design)
– 2 levels of sex (f vs. m)
– 3 levels of age (20’s, 40’s, 60’s)
• 6 different conditions (6 different people approaching concerns: sequence or carryover effects
– complete counterbalancing
– partial counterbalancing (Latin Square)
Mixed Factorial Designs
at least one factor (IV) manipulated between at least one manipulted within
e.g., add the factor of mood into above design
– (3x2x3 mixed factorial design) sex and age are manipulated within subject perform a between subject mood manipulation:
– Positive, Neutral, or Negative
concerns: equivalence for between subject factors concerns: carryover effects for within subject factors
IV x PV designs
At least one manipulated IV At least one subject variable
e.g., Personality Type (Type A or Type B) x competition (hi or low)
DV -- problem solving task
Main Effect of PV -- effect of person type Main effect IV -- task/situation effect Interaction -- task/situation differences depend on person type
– ** Must be careful in interpreting subject variables (as always)
Higher order designs
Three or more factors (IVs) Each level crossed with every other Each factor – may produce main effect May have interaction between any combination of factors
Males Caffeine Level
Low Medium High
Personality Type
A 40 30 20
B 40 50 60
Females Caffeine Level
Low Medium High
Personality Type
A 30 30 30
B 50 50 50
Higher Order Designs 2x3x2 Design Personality Type x Caffeine level x Sex DV: Problem solving speed
Males Caffeine Level
Low Medium High
Personality Type
A 40 30 20
B 40 50 60
Females Caffeine Level
Low Medium High
Personality Type
A 30 30 30
B 50 50 50
Higher Order Designs Main Effect of Sex
Males Caffeine Level
Low Medium High
Personality Type
A 40 30 20
B 40 50 60
Females Caffeine Level
Low Medium High
Personality Type
A 30 30 30
B 50 50 50
Higher Order Designs Main Effect of Caffeine Level
Males Caffeine Level
Low Medium High
Personality Type
A 40 30 20
B 40 50 60
Females Caffeine Level
Low Medium High
Personality Type
A 30 30 30
B 50 50 50
Higher Order Designs Main Effect of Personality Type
Interactions
Two Way Interactions– Personality Type x Caffeine Level
• (average across sex)– Personality Type x Sex
• (average across caffeine level)– Caffeine Level x Sex
• (average across personality type)
Three Way Interaction– Two-way interaction DEPENDS on level of other
variable
Three Way Interaction
0
10
20
30
40
50
60
70
Low Medium High
Type AType B
Males
0
10
20
30
40
50
60
Low Medium High
Type AType B
Females