3. The NELS data include a series of items (ByS44a to ByS44m) designed to assess students self-esteem
Question:
3. The NELS data include a series of items (ByS44a to ByS44m) designed to assess students’
self-esteem and locus of control. Conduct a CFA for the self-concept and locus of control items. First, check the descriptive statistics for the items in the raw data file
(ByS44a through ByS44m). Note that some of the items are worded positively and others negatively. Given the scaling (1 = strongly agree to 4 = strongly disagree), for positively worded items larger numbers actually represent worse self-concept or locus of control. I recommend you analyze the matrix data in the files “sc locus matrix.sav”
or “sc locus matrix.xls” where the positively worded items have been reversed so that for all items high scores represent better psychological health. Items that have been reversed end with an “r” in the matrix and in Figure 16.21. If you analyze the raw data I recommend you reverse these items.
Figure 16.21 shows the model that I recommend you start with. You will find that the model does not fit well (I obtained χ2 = 768.855, df = 64, and CFI = .780). First, examine the item wording (shown in Table 16.5). How might you modify the model to improve its fit (think of this as an informal theory method of revising the model)?
Do those modifications improve the fit to a statistically significant degree? Now take
a look at the modification indexes and the standardized residuals. Produce a table of correlation residuals (the sample correlation matrix minus the implied correlations).
What modifications might you make based on these various hints? How does the model fit now?
How many modifications did you make? Have you crossed the line from confirmatory analysis to exploratory analysis? Take a step back and think of your model more broadly. Might the model be better conceived as having more than two factors?
Might it be worth deleting some messy items? Discuss your models and your thoughts in class.
Step by Step Answer:
