Part III: Statistical Modeling

Over the past five workshops, we have learned how to load, clean and manipulate, visualize, and do some preliminary analysis on our data. Up until now, we have done very little actual statistical analyses. But isn’t the whole point to analyze our data? Yes, it is–and analyzing our data is less meaningful, less robust, and less informative if we don’t do the equivalent to our 5 workshops of work first.

The big picture: Stats is baking a pie

Think of it like baking: say we want to bake a pot pie. We’ll use the process of making the pot pie to represent analyzing our data. What do we have to do before you bake the pot pie? Well, we have to get your ingredients (aka load in data) and we have to prep those ingredients: dice chicken, cut up carrots, etc. (aka manipulate data). Sometimes before we bake, we have to do a little cooking, too, to best prepare ingredients–for example, we may need to cook our chicken on the stove before baking it in the pie (aka preliminary analyses). Last, but definitely not least, we have to make sure the pie looks like what we want/expect it to (aka visualizing to see what our data looks like). Only then can we bake the pie. And of course, we can’t just bake haphazardly–we have to know how to bake it: temperature and time (aka know what models to use).

Where we are at now

We’ve done all the prep. We’re now at the baking stage. For the next four workshops, we are going to be going through some common analytic methods, when to use them, how to compute them in R, and how to visualize results from each of those models. Here’s what we’ll look at:

  1. Group difference testing
  2. Correlation and regression
  3. Mediation and moderation
  4. Non-parametric regression

Themes of this section

One important thing we will learn, along with how to compute and interpret these tests in R, is how to choose the test to analyze your data. After all, if we don’t know when to use each test, there’s not much use in knowing how to use them. We’ll go into this in more depth, but there are 3 main things to consider when choosing how to analyze your data:

  1. What result do you want?
  2. What type of data is your dependent variable?
  3. Once finding the test that will get you that result and works with that dependent variable, are the assumptions met?

Over the coming workshops, we will use these three questions to explore and define when and how to use each model we discuss.