Goals This course is designed to (a) teach you the basics of classical factor analysis and path analysis as a basis for learning structural equation models, (b) introduce you to the distinction between exploratory and confirmatory models and (c) teach you structural equation modeling (SEM). After completing the course, you should be able to read most substantive applications of SEM, read some of the technical literature and apply the methods to your own research. Like any statistical technique, what matters most is experience and practice rather than sitting through lectures. If you want to learn this material, you have to actually do it.
Pre-reqs You need a decent background in regression analysis. Knowledge of factor analysis is not required, nor is a knowledge of matrix algebra. However, all of the basic literature in the field uses matrix algebra and the matrix notation introduced by Jöreskog. There are several versions of SEM programs which do not require matrix algebra, and we will use one of them at the beginning of the course, but if you do not learn the basic notation you will not be able to read the vast majority of published work, including hundreds of substantive papers.
Grades This is a seminar. You will be asked to do a set of ungraded exercises as a means of keeping up with the material. A paper applying SEM models to a substantive problem is required and will be the sole basis for your grade. You will present your papers during the last couple of weeks of class. Details on the paper will be provided later.
Computing To start, we will use AMOS, a "simple" program for SEM's. The program will is available on the lab computers, and you will also be given a free demo version, with limited capacity, to use on your own machine. After the learning the basics via AMOS, we will switch to industrial strength LISREL. More details on this will be provided later.
Labs The timetable lists a lab as part of this course. We will meet at a computer lab for a couple of regularly scheduled course sessions, and I will schedule other lab hours as needed.
Copy Fees A good deal of additional printout and other material will be made available as we go along. In order to defray departmental costs, a $10.00 fee will be charged. Please make out checks to the University of Illinois and give them to me.
Office Hours Monday 12:00- 1:00PM and by appointment. "By appointment" means what it says; I will be hard to find on a walk in basis, but I am more than willing to schedule a time to see you. My BSB office phone: 413-3759; E-mail: DCAMP@UIC.EDU. I much prefer the latter. You will find that I answer e-mail religiously and detest the phone. You can send me outputs to look at via e-mail. I will cover details of how to do this in class.
Texts: (1) Arbuckle, James: Amos Users' Guide. 1996. Smallwaters Corp: Chicago. (ARB)
(2) Maruyama, G. G. Basics of Structural Equaition Modeling.
| Week | Date | Topic | Readings & Assignments |
| 1 | Aug. 21 | Introduction/Regression Review | |
| 2 | Aug. 28 | Classical Path Analysis | |
|
|
|||
| 4 | Sept. 11 | Classical Factor Analysis | |
| 5 | Sept. 18 | Confirmatory Factor Analysis I | |
| 6 | Sept. 25 | Confirmatory Factor Analysis II | |
| 7 | Oct. 2 | Model Selection | |
| 8 | Oct. 9 | Mathematics of CFA | |
| 9 | Oct. 16 | Multiple Group Comparisons | |
| 10 | Oct. 23 | Miscellaneous Topics | Catch up on Reading1 |
| 11 | Oct. 30 | Structural Equation Models I | Paper Proposal Due |
| 12 | Nov. 6 | Structural Equation Models II | |
| 13 | Nov 13 | LISREL notation and matrices | |
| 14 | Nov. 20 | More on LISREL | |
| 15 | Nov. 27 | Paper Presentations | |
Details on Course Paper
Your major activity in this course will be to write a paper using some aspect of confirmatory factor analysis and/or structural equation models. In order to make sure that you are not embarking on something that is far too difficult or impossible, I am asking that you give me a brief one page proposal by Oct. 16. I will, of course, be happy to discuss topics with you before that. Your paper does not have to be what one normally thinks of as a "seminar paper" in the sense that it is a full blown attack on a particular substantive problem. It does have to demonstrate that you can define a problem, do an appropriate analysis and interpret the results. You do not have to demonstrate that no one has ever done this before or that your approach is better than anything else out there. Since, as you will learn, SEM work can be done off published correlation matrices, it is often interesting to re-analyze published data. In answer to the inevitable question: "Is just a confirmatory factor analysis enough," the answer is yes, but you will need to demonstrate that you can select an appropriate model from competing alternatives. All of this will make more sense to you in a few weeks. In the meantime, start thinking about data sets that you might like to use.