PSY 350 Lecture Notes - Lecture 59: Analysis Of Covariance, Analysis Of Variance, Dependent And Independent Variables
PSY 350 5-2-18
Wednesday, May 2, 2018
9:05 AM
Testing Moderation
-look for interactions
-we know how to do this in ANOVA
-you can find an interaction between a continuous variable and a grouping variable in
ANCOVA
-you can find an interaction between two continuous variables in multiple regression
-remember there are three variables in an interaction, or a moderating relationship
-don't forget the dependent variable
-the moderator and the IV/predictor do not influence one another
Challenges in Finding Moderators
-moderating effects are often small
-need large samples to find statistically significant effects
-if we find them, are they practically significant?
-choosing which variable is the focal variable vs. the moderator is sometimes messy
-different researchers could focus in different places
-as you know, interactions can be difficult to interpret
-lots of forms a moderating relationship could take
-strengthening/reducing an effect
-changing the direction of an effect
-eliminating an effect
Mediation
-if moderation is about when one variable affects one another, mediation is about how
Causally situated between X and Y means that M comes between X and Y
How mediation works
-a mediating variable explains how we get from one variable to another
"…because…"
"…due to…"
"…by…"
Like links in a chain--if we break the link from X to M, then Y doesn't happen
-different from a moderator: the relationship between X and Y doesn't change at
different levels of M
-Rather, X influences M, which then influences Y
-M is not an independent variable
find more resources at oneclass.com
find more resources at oneclass.com
Document Summary
We know how to do this in anova. You can find an interaction between a continuous variable and a grouping variable in. You can find an interaction between two continuous variables in multiple regression. Remember there are three variables in an interaction, or a moderating relationship. The moderator and the iv/predictor do not influence one another. Need large samples to find statistically significant effects. Choosing which variable is the focal variable vs. the moderator is sometimes messy. As you know, interactions can be difficult to interpret. Lots of forms a moderating relationship could take. If moderation is about when one variable affects one another, mediation is about how. Causally situated between x and y means that m comes between x and y. A mediating variable explains how we get from one variable to another. Like links in a chain--if we break the link from x to m, then y doesn"t happen.
