Structural Models in SEM
Readings
Be prepared to answer the following questions from Schumacker and Lomax in class. Page numbers for answers are given. Feel free to skim other sections of the book and concentrate on these critical questions. We will be concentrating on Amos, though the authors also discuss LISREL and EQS packages.
CHAPTER 7: PATH MODELS
1. Differentiate path models from regression or the usual type of SEM model. p. 150
2. In Figure 7.1, why are there three error terms even though there are five observed variables? p. 152
3. In Figure 7.1, why is there a curved covariance arrow connection only two of the five variables? p. 152
4. In path modeling, what is the symbol for an observed variable? A latent variable? An error term? p. 153
5. Since both regression and path analysis compute regression coefficients, what's the difference? p. 155
6. What are path coefficients? p. 158
7. Using the example on p. 158, what is decomposition of (partitioning) the correlation about in path analysis? p. 158
8. What is full information estimation? What is the relevance to path modeling with AMOS compared to
SPSS regression procedures? p. 159
9. When you do model testing in path analysis using a SEM package, do you get model fit measures the same
as if you were doing SEM with latent variables? p. 160
10. What are modification indexes (MI) used for in path analysis? pp. 161-162
11. Why look at standardized residuals > 1.96? p. 161
CHAPTER 9. DEVELOPING STRUCTURAL EQUATION MODELS, PART I
1. Why are latent variables supposed to be measured by three or more observed/indicator variables? p. 198
2. What is an error variance term? What is the error variance when the reliability is known? How is it entered in Amos? pp. 199, 208
3. What are "validity coefficients"? p. 199
4. What is the implied covariance matrix and what is its role in SEM? p. 208
CHAPTER 10. DEVELOPING STRUCTURAL EQUATION MODELS, PART II
1. What is the example model in this chapter about? pp. 214-215
2. Why do only some of the arrows from the latents to their indicators have a "1" on top? pp. 214-215
3. Which are the structural arrows, for which structural (path) coefficients to be estimated? p. 215
4. What is the relation of Tables 10.1 and 10.2 on pp. 221-222?
5. Why is the initial model rejected? p. 223
6. How is the respecified model different, in Figure 10.2, p. 225?
7. Should one always drop non-significant structural coefficients (paths) from the model? Pp. 225-226 and Table 10.3
8. Is the respecified model acceptable? Why or why not? pp. 223, 226
CHAPTER 11: REPORTING SEM RESEARCH
1. When reporting SEM research, what do APA guidelines require? p. 231
2. What are the 10 commandments from Thompson (2000) for reporting SEM research? pp. 231-232
3. Do Schumacker & Jones recommend including the Amos, Lisrel, or EQS program in published articles? p. 236
4. In Table 11.1, what coefficients are reported, apart from the factor loadings? p. 241
5. What is the measurement invariance test? pp. 242-243
6. For what was the measurement invariance test a prerequisite? p. 243
Homework
Be prepared to answer the following questions from Schumacker and Lomax in class.
HOMEWORK CHAPTER 11: REPORTING SEM RESEARCH
1. Load the file Fig11.2.amw from the Chapter11 folder of the textbook CD.
2. From the menu, click File, Data Files, to confirm this has loaded girls_cor.sav as the data file.
3. In the Data Files dialog, click on the View Data button and confirm the data look as shown on p. 249 of the text. Afterwards, click OK.
4. Look at the graphical model you have loaded and confirm it looks exactly like Fig. 11.2. Use the Erase Objects tool (the "X" icon)
to erase the arrow from the Spatial latent variable to the Lozenges indicator variable.
5. Run the model as in the previous homework exercise, and print it out in the same manner.
6. Be prepared to explain why this model is not acceptable.
7. Redraw the arrow from Spatial to Lozenges, using the Draw Paths tool (single-headed arrow icon)
8. Next we wish to run Specification Search to find an alternative, better model. Note the Amos interface is a little different from the textbook now.
Select Analyze, Specification Search (not Model Fit, Specification Search).
9. In the Specification Search dialog box, click the Make Arrows Optional icon (the firs, left-most icon). Then click on the arrow from
Spatial to Lozenges to make it optional. It turns yellow.
10. Also make the path from Verbal to Lozenges optional. (Note you can make several arrows optional by holding the left mouse
button down and dragging the mouse pointer through the arrows.)
11. In the Specification Search dialog box, click the Perform Specification Search button (the VCR-type play button).
12. In the Specification Search dialog box, click the Printer icon to print out the results, which will include four possible default
models, with the best showing in boldface.
13. Have the main Amos window (the diagram) below the Specification Search window. Double-click on each of the four default
models to note which is which, as the arrows change in the main window.
14. Using AIC, be prepared to rank the four models by model fit.
15. Before leaving Specification Search, set the optional arrows back to not optional (use the second-from-the-left icon).
16. Back in the main window, use the erase (X) tool to make the diagram look like the best model by specification search.
17. Run the model as before, printing the output, and be prepared to interpret it in class.
Explore