Before performing the structural equation modeling (SEM) procedure, a CFA was used to determine if the proposed measurement model specified the posited relationships of the observed variables to the latent constructs. A second-order factor structure was used for nostalgia to explore the tree-like factor structure where the factors of one stage become variables of another. Results of the CFA revealed an acceptable model fit with the exception of the chi-square test, χ2(336) = 602, p < .001. However, since the chi-square test is sensitive to larger sample sizes, other model fit indices have been suggested to provide better evaluation of model fit (Bentler 1990). Since the other goodness-of-fit indices were good, GFI = .94 (good if greater than .90), TLI = .93 (good if greater than .90), CFI = .94 (good if greater than .90), RMSEA = .05 (good if less than .05) (Hair et al. 2010; Marsh, Hau, and Wen 2004), the model fit was deemed to have satisfactory fit. Construct reliability was evaluated by estimating composite construct reliability (CCR) (Hair et al. 2010). Since all constructs had CCR values greater than the criterion (.70), the measures were deemed reliable (Fornell and Larcker 1981; Hair et al. 2010). Validity was tested by examining the t-values between each construct and their latent variable and the average variance extracted (AVE) for each construct. All 28 items in the measurement model had significant (p < .05) t-values on their respective latent constructs, which suggests the measurement scales for each construct had a high level of convergent and construct validity. In addition, the AVE values for each construct were greater than .50, suggesting a high level of convergent validity. Discriminant validity shows the extent to which a given construct differs from other constructs (Hair et al. 2010). One way to evaluate discriminant validity is to determine if the AVE values for each construct are greater than the squared correlations between the construct and all other constructs (Fornell and Larcker 1981). Since the AVE values were all greater than the squared correlations, the measurement model was also deemed to demonstrate good discriminant validity. The results are indicated in Table 3.