If you thought that the previous chapter was bad, well, I'm about to throw an added complication into the mix. We can combine repeated-measures and independent designs, and this chapter looks at this situation. As if this wasn't bad enough, I'm also going to use this as an excuse to show you a design with three independent variables (at this point you should imagine me leaning back in my chair, cross-eyed, dribbling and laughing maniacally). A mixture of between-groups and repeated-measures variables is called a mixed design. It should be obvious that you need at least two independent variables for this type of design to be possible, but you can have more complex scenarios too (e.g., two between-groups and one repeated-measures, one between-groups and two repeated-measures, or even two of each). SPSS allows you to test almost any design you might want to, and of virtually any degree of complexity. However, interaction terms are difficult enough to interpret with only two variables, so imagine how difficult they are if you include four. The best advice I can offer is to stick to three or fewer independent variables if you want to be able to interpret your interaction terms,1 and certainly don't exceed four unless you want to give yourself a migraine.
This chapter will go through an example of a mixed ANOVA. There won't be any theory because you've probably had enough ANOVA theory by now to have a good idea of what's going on (you can read this as 'it's too complex for me and I'm going to cover up my own incompetence by pretending you don't need to know about it'). Essentially, though, as we have seen, any ANOVA is a linear model, so when we have three independent variables or predictors we simply add this third predictor into the linear model, give it a b and remember to also include any interactions involving the new predictor. We'll look at an example using SPSS and spend a bit of time developing your understanding of interactions and how to break them down using contrasts.
1 Fans of irony will enjoy the four-way ANOVAs that I conducted in Field and Davey (1999) and Field and Moore (2005), to name but two examples.