What is a cross-over interaction effect in a two-way ANOVA?
Where the effect of one IV flips or reverses depending on the level of the other IV
How can you remove a cross-over interaction effect?
Sometimes you can switch the IVs in their positions on the graph to remove the crossover effect
How do you write a conclusion for an ANOVA with a cross-over interaction effect?
You cannot specify direction because you have only done one statistical test and there are technically two directions A1 less than A2 in condition B1, A2 less than A1 in B2.
So the conclusion will be vague, but accepted in PSYC3010: “The difference between pain education and no pain education for exercise control is in the opposite direction to the difference between pain education and no pain education for pain self-management, and the differences in these differences is significant (stats here)”.
This is vague because there is no indication of direction
What is a simple main effect?
Where we look at the effect of one independent variable on the dependent variable but separately for the levels of the other independent variable, ie. looking at the differences rather than the differences of the differences
Why do we look at simple main effects?
It can help us clarify an interaction effect
What is the difference between normal contrasts and simple main effect contrasts?
Simple main effect contrasts compare the effects of one IV within one level of the other IV, meaning that there are no families in these contrasts.
This also means that the contrasts are likely not orthogonal. Hence we cannot use DER
What is the difference between a simple main effect and an interaction effect?
A simple main effect looks at the differences between two groups one level of an IV. An interaction effect looks at the differences between two groups and compares this to the differences of two groups across the levels of an IV.
A simple main effect is about differences, an interaction is about the differences between the differences
How can simple main effects contradict with an interaction effect?
If you have a significant simple main effect for one comparison, but not another comparison, you would say the differences between the groups is different - which is an interaction effect.
However, there may not be a significant interaction effect. This is contradictory but it happens because of type I error control
Can you have more than two-way ANOVA?
Yes you can, but it gets complicated. Just do MLR <3