CHAPTER 1: INTRODUCTION
1. Differentiate latent and observed variables. p. 3.
2. Identify and explain the four types of modeling which can be implemented by SEM. p. 4.
3. What four reasons do the authors give for the popularity of SEM? p. 7.
CHAPTER 2: DATA ENTRY AND DATA EDITING ISSUES
1. Describe data entry using AMOS. pp. 14-17.
2. What is the data issue pertaining to measurement scale? p. 24.
3. What is the data issue pertaining to restriction of range? p. 25.
4. What is the data issue pertaining to missing data? How does AMOS handle missing data? pp. 25-6.
5. What are the data issues pertaining to outliers and nonlinearity? pp. 31-33.
6. What is the data issue pertaining to non-normality? p. 33.
CHAPTER 3: CORRELATION
1. Why is correlation being discussed in depth in this chapter?
2. What types of correlation coefficients are used in SEM, and when? pp. 39-40.
3. If you are told the "covariance matrix is non-positive definite" and a SEM solution
is not possible, what might be the cause? p. 48.
4. What is the independence model in SEM? The saturated model? p. 49.
5. How large a sample is needed in SEM? p. 49.
6. Why does SEM use covariance rather than correlation matrices as input? p. 55
7. Should data be standardized? pp. 55-56
CHAPTER 4. SEM BASICS
1. What is specification error and why is it important in modeling? pp. 62-3.
2. Do we want underidentified, just-identified, or overidentified models? p. 64
3. What is the recursivity issue in modeling? p. 65
4. What is estimation? What is its purpose? What are the major types, and pros and cons of each? pp. 66-68.
5. How do you know if your data are normal enough to use the usual ML method of estimation,
or if you have to use WLS or ADF instead? p. 69.
6. What are the two major types of model testing in SEM? pp. 69-70.
7. What is the "modification index" and what is it used for? pp. 72.
8. What is "specification search"? p. 73
9. Why look at the standardized residual matrix? p. 74
10. What is cross-validation and how does it relate to SEM? p. 74
CHAPTER 5: MODEL FIT
1. What are "model fit" coefficients in SEM? (No specific page).
2. What is the model chi-square test and what are its limitations? pp. 82-83.
3. Interpret Table 5.2 on p. 87. Is this a good model? pp. 87-88.
4. Which four goodness of fit measures are used to compare SEM models with the null (independence) model? p. 103.
5. What is "model parsimony"? How large should parsimony coefficients like PNFI be in a good model? pp. 104-105.
6. AIC does penalize for lack of parsimony, so how is it different from PNFI? p. 105
7. What is the two-step modeling approach? p. 106
8. What is the four-step modeling approach? pp. 106-107
9. What is "testing for measurement invariance"? p. 108
10. Should you grow the model using modification indexes, then pare the model by deleting non-significant parameters (paths)? Or the reverse? p. 109.
11. Differentiate the likelihood ratio (LR), Lagrange multiplier (LM), and Wald tests of parameter estimates. pp. 109-110.
12. In Amos, the Specification Search option is an automatic way of generating and testing a set of related models.
In the Specification Search dialog box at the bottom of p. 110, why is Model 2 the best?
13. Which model fit index seems relatively unaffected by sample size? p. 115
CHAPTER 6: REGRESSION MODELS
1. What is the purpose of this chapter? (No specific page).
2. What is specification error? p. 129.
3. What three things make Figure 6.1 a regression model? p. 130.
4. Why is Figure 6.1 a saturated model? p. 130
5. Interpret the formula under Section 6.5. p. 130
6. Why do model fit statistics, usually central to SEM, not apply in a SEM regression model? p. 135
7. What do the authors cite as the three main problems of regression analysis? Why is SEM better for each? p. 138
8. What are the main steps in AMOS for getting regression output? pp. 138-140
CHAPTER 8: CONFIRMATORY FACTOR ANALYSIS
1. What is the major limitation of path models, according to the authors? p. 168
2. What does a confirmatory factor analysis model consist of in terms of variables and relationships (arrows)? pp. 168-169
3. How could we use CFA to test for correlated error among indicators? p. 170
4. Does CFA tell us how many latent variables to have and/or which indicators go with which latents? pp. 171-172.
5. In Table 8.1, bottom section for the standardized residual matrix, what are these coefficients? Why are two in boldface? p. 175
6. By what criteria is the model in Figure 8.2 (p. 174, the :"misspecified model") rejected? pp. 176-177
7. What coefficients are used to modify CFA models? p. 177
8. What coefficients are used to assess the reliability of a CFA model? p. 180
HOMEWORK CHAPTER 5: MODEL FIT
1. Load FACTOR.AMW from the textbook disk, Chapter5 folder.
a. Extra credit: draw it manually!
2. From the menu select File, Data Files; then in the Data Files dialog, click on View Data to verify that grant.sav has been loaded
and to view it in the SPSS data editor.
3. From the menu select View, Analysis Properties, and select the output options shown on p. 86 (Minimization history,
Standardized estimates, Squared multiple correlations, Modification indices). Also select Factor Score Weights.
4. In the Analysis Properties dialog, click the Estimation tab and verify you are using Maximum Likelihood estimation.
5. In the Analysis Properties dialog, click the Title tab and insert your own name and "Homework Chapter 5".
6. From the menu, select Analyze, Calculate Estimates.
7. From the menu, select View, Text Output.
8. Click the options icon (the checkbox symbol), select the View tab, and check "View entire output file".
9. Click the print icon and print the output.
10. Click on the "View the output path diagram" icon (right under "Tools" in the menu). Select File, Print; check you
want standardized estimates for the default model; print the path diagram.
11. Annotate the print output manually and be prepared to discuss it in class, section by section.