When you understand the meaning of Linear Regression Coefficients for All Types of Predictor Variables--Dummy Variables, Interactions, Quadratic, Centered and Standardized Variables--you can build more sophisticated and more accurate models and easily translate your findings into meaningful results.

Whether your statistical training focused on linear regression or ANOVA, there are usually gaps in that training. Most linear regression courses assume all your predictor variables are continuous, and ANOVA courses assume that all predictor variables are categorical.

But whether you primarily do observational or experimental research, at some point you are going to need to build more than a simple model with one type of predictor.

The beauty of the regression model is that it can incorporate all kinds of predictors:

• Both binary and multicategory dummy variables
• Interaction terms to test if effects differ in different situations or populations
• Quadratic terms that allow you to test a non-linear relationship with a linear model
• Centered variables that make interpretation of coefficients easier, without changing distributions
• Standardized predictors that remove the effects of the units of measurement

And when you understand how to interpret the effects of these predictors in a linear regression model, you can easily interpret the same kinds of effects in more complicated models, such as logistic regression or multilevel models. All these kinds of predictors work the same way in any kind of regression model.

So if you need to learn more complicated regression modeling, make sure you master this first. Then when you go on to learn multilevel models, logistic, Poisson, or any other kind of regression, you can focus on the unique challenges in that type of modeling, and not get stuck on the foundation.

Module 1: Introduction and Review of Regression Models with Continuous Predictors

• A Review of the Simple Linear Regression Model
• A Review of the Multiple Linear Regression Model with continuous predictors
• Interpreting Regression Coefficients when Predictors are Correlated

Module 2: Scaling of Dependent and Independent Variables

• Understanding Unstandardized and Standarized Coefficients
• How and when to center predictors to improve interpretability of coefficients
• How transforming Y or X to meet assumptions affects the interpretation of Coefficients
• When centering affects the meanings of other coefficients in the model.

Module 3: Multiplicative terms for Continuous Predictor Variables

• How including polynomial terms changes the meaning of other coefficients
• An introduction to interactions
• Interpreting interactions between two continuous variables
• Interpreting interactions with quadratic terms--changing the curve

Module 4: Dummy Coding Binary Predictors: Main Effects and Interactions

• The basic difference between dummy and effect coding
• How to dummy code binary predictors, what the coefficients mean, and when including dummy variables changes other coefficients
• How to interpret interactions for dummy-coded predictors
• How including interactions changes coefficients of main effects (this is different from ANOVA!)

Module 5: Dummy Coding Multi-Category Predictors: Main Effects and Interactions

• How to dummy code multicategory predictors and interpret their coefficients
• How to interpret interactions for multicategory dummy predictors
• How to interpret interactions between two dummy-coded variables

Module 6: Putting it all together

• Now that you understand all the pieces, we'll see how all the terms fit together in one big model, and how to break the model apart to understand each term. (This is often the trickiest part).

• Learning an extension of linear regression (such as multilevel, logistic regression, or proportional hazard models) and want to reinforce your current understanding.  Many of the concepts covered in this workshop are vital in learning extensions to linear models.  You’ll be able to progress to the next level much faster and easier if you have an intuitive, gut-level understanding of these concepts.
• Using linear regression or Analysis of Covariance for the first, second, or fifth time.
• A Graduate student who needs to defend your regression analyses, and needs a rock-solid understanding of the meaning of your models.
• Are a researcher in Psychology, Education, Agriculture, or another field in which your statistics training focused on ANOVA, and not so much on linear regression.  You’re finding your lack of understanding of regression, and where ANOVA fits into it, is limiting your ability to apply appropriate analyses to your data.

Prerequisites:

• You will get the most out of the workshop if you have had at least two statistics classes—an intro class, and a class that contains multiple regression or Analysis of Variance (ANOVA).  Beginners will have trouble keeping up.  That said, if you’ve had one statistics class, but are working on regression and are confused, this workshop will be helpful. 
• You should be familiar with SAS or SPSS.  You are free to use another program if you prefer, but I will only support SAS and SPSS in examples.  I have used, but am less familiar with JMP, Stata, and SPlus, and I will answer questions about them if I can.