Because I have a medical appointment on Thursday, and must miss class, I am putting these notes together in a way that I hope will be meaningful as a stand-alone.
This lecture is really a PowerPoint presentation. This page provides the link to that presentation. You probably have PowerPoint on the machine that you are using, so double click on the link given below.
Also look at the web page that I put together.
So far I have talked about the basic structure of multiple regression, and shown an example. In the lab students went through the height and weight example, and saw some important issues in multiple regression.
The basic theme in all this is on how we put models together and create a meaningful analysis of a set of data. Note that the emphasis is more on "meaningful" than on "most predictive." Often a model with a smaller R might still be more important to the experimenter than a model that has a higher R but doesn't fit together logically.
Today I want to give another example, and look at hierarchical regression.
Much of the emphasis will be on what these particular data have to tell us.
We will look at
- full and reduced models
- squared semi-partial correlations
- tolerance
- Interactions
- Moderating and Mediating relationships
What I say here will come directly from the "paper" I put together on the web.
The rest of the class is based heavily, though not entirely, on that paper. You can get it by going to the link. (That paper was written using an older version of SPSS. The fact that the tables will look a tad different is really not important. They carry all of the important information.)
You can download the presentation itself, but the one that you will find is written for Office 97, but it should run on any more recent version of PowerPoint.
Last revised: 03/24/02