Chapter 9: Experiments with more than one independent variable Flashcards
(15 cards)
Factorial designs
When at least one of the independent variables is manipulated
Tells us testing boundaries of effects + testing theory
2x2 factorial design
Two independent variables that each has two levels
Independent variables are crossed with each other so that we get four conditions
Variables can be manipulated and/or measured and may be within-subject or between-subject level
2x2x2 design: add another independent variable with two levels
3x3 design: two independent variables with three levels each
Interaction effect vs no interaction effect
Interaction (= moderation): effect of variable A depends on the value of variable B
No interaction: effect of variable A is almost the same for group 1 as for group 2
Main effect
The average effect of one independent variable on the dependent variable, across different levels of the other independent variable(s)
By calculating these averages, we calculate and compare the marginal means
For each independent variable, you have one main effect that you can test
Interaction effect
The effect of one independent variable on the dependent variable, conditional on the value of another independent variable
Typically more important than main effects in factorial designs
How to recognize main effects
Main effect of photo type: calculate the marginal mean of the group that was shown the alcohol photo and the group that was shown a plant photo → if there is a significant difference between these two, there is a main effect
How to recognize two-way interaction effects in data
Does the effect of photo depend on the type of word you see:
1. Compare the difference between the values of photo type
2. Look whether there’s a difference between the differences: if the differences aren’t equal, there is an interaction effect
3. Check if the difference is statistically significant (data in article)
How to recognize two-way interaction effects in figures
If the lines run parallel, there is no interaction
If the lines are not parallel, there is an interaction
In a bar figure: height of the bar indicates the average score of the independent variable → connect the bars that are related to each other and check if lines are crosses or parallel
Describing main effects
Two independent variables: language lessons or not, language test or maths test
Dependent variable: score on test
If there is a main effect of lessons: ‘There is a main effect of lessons, such that the group that got the language lessons scored higher than the group that did not receive lessons.’
If there is no main effect of test type: ‘There is no main effect of test type.’
Describing interaction effects
We only look at one level of the other independent variable: we don’t look at the effect of the first independent variable on the second independent variable (horizontal in stead of vertical)
→ Look at simple main effects and describe what you see
Extra things to take into account when variables are manipulated at the within- or between-person level
Implication for numbers of participants: if you want 12 participants in each condition, this changes the sample size for each design
- Independent groups or between subjects: 12 people needed in each condition (4) → 48 people needed
- Within groups: only 12 people needed because every person goes through all the conditions
- Mixed designs: one variable is manipulated at between-subjects level and the other at between-subjects level → 24 people needed because you are either in the upper two or lower two groups
- Implication for statistical tests: type of test will differ because standard ANOVA is not correct for mixed design
2x2x2 factorial design: market research on chips
Dependent variable: preference for type of chips
Independent variables: two types of chips (salt/paprika), two brands (well-known/store brand), two age groups (children/adults)
How to visualise: two crosstables or two figures (one for paprika and one for salt)
Two-way interaction effect vs three-way interaction effect
Two-way: interaction effect between two independent variables
Three-way: interaction effect between three independent variables (idea: check for a difference in the difference between the differences)
How to recognize three-way interactions effects in figures
- Both patterns are parallel → no three-way interaction
- Two two-way interactions that are not identical (look different) → type of interaction depends on third variable → three-way interaction
- Two two-way interactions that go in the same direction (look the same) → no three-way interaction
- Two two-way interaction that look similar, but bigger from each other → three-way interaction
! If the first figure doesn’t have a two-way interaction and the second one does (or the other way around), they still differ so there is a three-way interaction
2x3 factorial design
The first independent variable only has two levels
The second independent variable has three levels
Finding one difference between the three differences or one line that is not parallel to the others is enough to say that there is an interaction effect
