Most psychologists are highly familiar with the analysis and interpretation of interaction effects in factorial analysis of variance (ANOVA). In ANOVA these interactions are between categorical independent variables (e.g., experimental conditions). What many researchers do not realize is that interaction effects in ANOVA are a special case of interaction effects in multiple regression which can be computed in much the same way. A common, but unnecessary and inappropriate, strategy for dealing with interaction effects between continuous independent variables is to turn them into categorical ones (e.g., by a median split). A better approach - one with greater statistical power and which is more informative - is to add interaction effects to a multiple regression model. This approach is often known as moderated multiple regression. This workshop will: introduce moderated multiple regression, explain how to compute an interaction term for moderated multiple regression, discuss the role of centring in making the results easier to interpret, and demonstrate how to run and interpret such a moderator analysis in SPSS.
You can view the presentation on google docs:http://docs.google.com/TeamPresent?fs=true&docid=dp86h6t_3wmr5ft
The workshop is supposed to be a basic introduction. For a more advanced introduction I'd recommend Kris Preacher's web page:
A primer on interaction effects in moderated multiple regression.
Kris Preacher also has links to excellent tools for investigating and plotting moderator effects. (These generate R code that you can plot online with one mouse click or paste into your R workspace to edit for publication quality graphics). Although, for initial exploration it may be quicker to use the regression equation to plug in different X values and see how the expected Y value changes. I'd probably do this by hand (or maybe Excel) and sketch out basic plots by hand.
No comments:
Post a Comment