Analysis of Variance in R: A Multipart Series

Like most applied statisticians, I’ve grown into my role of statistician from a major field that was not statistics–it was psychology (experimental psychology with a dash or behavioral neuroscience to be exact). As such, the first tool of the trade when it came to analytical software was SPSS. (It was actually SPSS for DOS if you can believe that.)

As I’ve grown in my career, and helped more and more clients, I’ve had the opportunity to learn more and more software. And one thing typically comes up when clients come in using R (especially if they are coming from a software like SPSS)…ANOVA and the Type III Sums of Squares.

Now, the question of “which type” of Sums of Squares to use has been asked (and answered) in several places. See HERE, HERE, and HERE for starters. But, when a client wants to use R as a replacement for SPSS or SAS (or when I wanted to replace SPSS with R), I really just need to know how to accomplish that task: type III Sums of Squares for an ANOVA in R. And I’ll show you how below.

Over the next few weeks, I’ll be posting about running (set-up, analysis, assumption checks, graphing, interpretation) various kinds of ANOVA models. I’ll cover:

• One-way between
• One-way within
• Two-way mixed design (within and between)
• Two-way between

These are the most common types of ANOVA models that I see in my consulting role. They also have the highest pain-points for clients, when they are using R. I’ll show you the correct way to run all these ANOVA models–specifically in R. And by “correct” I mean the most common type of ANOVA–the Type III Sums of Squares models.

See you soon!