1/50
Looks like no tags are added yet.
Factorial design
A design testing the effects of more than one independent variable, including their main effects and interactions.
Factor
A variable whose effects are examined in a factorial design.
Types of factorial designs
Factors can be implemented between subjects, within subjects, or in a combination of both.
Reasons to use factorial designs
Test interactions, investigate multiple IVs efficiently, and account for one variable while assessing another.
Main effect
The overall effect of one IV on the DV, averaging across the levels of the other factors.
Interaction
When the effect of one IV on the DV depends on the level of another IV.
Main effect versus interaction
A main effect describes an IV's overall relationship with the DV; an interaction describes how that relationship changes depending on another IV.
Interpreting main effects with interactions
A main effect may hide important differences across levels of another factor, so interpretation must consider the interaction.
Two-way interaction
An interaction involving two IVs.
Three-way interaction
An interaction in which the two-way interaction between two IVs changes across levels of a third IV.
Conditional effect
The effect of an IV at a particular level or value of another IV.
Joint effect
The combined relationship of IVs with the DV; the slides use this phrase when discussing interactions.
One-way design
A design with one IV; it does not test an interaction between different IVs.
Factorial notation
Each number gives the number of levels in one factor; the number of numbers gives the number of factors.
2 Ă— 2 design
Two IVs with two levels each, producing four combinations in a fully crossed design.
2 Ă— 3 design
Two IVs, one with two levels and one with three, producing six combinations.
2 Ă— 2 Ă— 3 design
Three IVs with two, two, and three levels, producing 12 combinations.
3 Ă— 2 Ă— 4 design
Three IVs with three, two, and four levels, producing 24 combinations.
Calculating factorial conditions
Multiply the number of levels of all factors to find the number of combinations in a fully crossed design.
Interactions across measurement levels
Interactions can involve categorical variables, continuous variables, or a combination.
Interaction example
The effectiveness of an antidepressant differs depending on the type of therapy used.
Symmetry of an interaction
An interaction can be described as the effect of A depending on B or the effect of B depending on A.
Independence of effects
In the lecture's experimental examples, factor assignments should allow the separate effects of the IVs to be distinguished rather than confounded.
Confounded factors
Factors vary together in a way that prevents their separate contributions from being identified.
Orthogonality
Independence between factors or effects; in the lecture, knowing a person's level of one factor should not determine their level of another.
Orthogonal design
A design arranging factors so that their effects can be distinguished without confounding their assignments.
Crossed design
A design pairing every level of each IV with every level of every other IV.
Crossing therapy with medication
Include every therapy–medication combination so the effects can be compared across the levels of both factors.
Non-crossed design
A design omitting some combinations of factor levels, potentially making effects difficult to separate.
Power benefit of factorial designs
Relevant additional factors can explain outcome variance, reducing unexplained error and improving detection of effects.
Unexplained variance
Variation in the DV not accounted for by the factors and interactions in the analysis.
Efficiency of factorial designs
Several effects can be tested in one study, and reduced error can lower the sample needed to detect main effects.
for testing main effects- factorial designs are most efficient
Factorial condition-count tradeoff
A 2 Ă— 3 design has six combinations, whereas separate two-group and three-group studies have five groups altogether; efficiency depends on more than group count.
Complexity of factorial designs
Additional variables and procedures create more possible sources of error and make problems harder to diagnose.
screwing up one can screw up all
Order effects in factorial designs
When factors are presented sequentially, presentation order may affect results or appear to create an interaction.
Controlling order in factorial designs
Counterbalance or model presentation order when appropriate; doing so adds further design complexity.
Variables versus constructs
Study variables operationalize constructs; independently arranging variables does not guarantee that the underlying constructs are independently changed.
Construct-confounding example
An alternative-partner manipulation may also increase gratitude toward one's current partner, making an apparent interaction difficult to interpret.
Measured versus manipulated factors
Measured factors may share causes and be correlated, making causal interpretations more difficult than with independently manipulated factors.
Third-variable problem in factorial designs
A third variable may influence both predictors and the outcome, producing an apparent interaction with another explanation.
Interaction effect size
Interactions are often smaller than main effects and can require substantially larger samples to detect.
Underpowered interaction test
A study with insufficient power may miss an important interaction even when its main effects are detectable.
Choosing a factorial analysis
Consider the measurement levels of the IVs and DV and whether the analysis can test the relevant main effects and interactions.
Chi-squared in factorial research
The slides discuss chi-squared approaches for nominal variables; analyses involving several categorical factors require care to represent the full design.
Factorial ANOVA
An analysis testing main effects and interactions when factors are categorical and the DV is interval or ratio.
One-way ANOVA
An ANOVA with one factor.
Two-way ANOVA
An ANOVA with two factors, testing their main effects and interaction.
Multiple regression
An analysis assessing how well a set of predictors explains or predicts an outcome.
Regression in factorial research
Can include continuous and categorical predictors and interaction terms to assess their relationships with an interval- or ratio-level outcome.
Moderation analysis
A regression approach testing whether the relationship between one predictor and the outcome depends on another predictor.
Factorial study-method example
Three study methods crossed with two study times form a 3 Ă— 2 design with six combinations.