# Here, Y. Denotes The Five Observed Indicators, Tau. Is

1200 WordsMar 31, 20175 Pages

Here, y denotes the five observed indicators, au is intercept for each of the indicators, lambda is the factor loadings, eta is latent variables and e is random measurement error for each of the indicators. Furthermore, since this model applies effect coding by not fixing the first loading for each factor to be 0, it is given that lambda_{1ij}+lambda_{2ij}+lambda_{3ij}=3 and lambda_{4ij}+lambda_{5ij}=2 .
When comparing groups, the multigroup confirmatory factor analysis (MGCFA) method is applied, thereby allowing for comparison of parameter estimates between countries. However, such comparison depends on the parameters being measurement invariant across all groups. As mentioned in previous studies (Ariely and Davidov, 2011),*…show more content…* In the results section, I am going to show the different measurement invariance models.
In order to determine whether national political trust (NPT) and supranational political trust (SPT) can be regarded as two empirically distinct concepts, it is necessary to test a one-factor model encompassing all five indicators against the proposed two-factor model shown in Figure 1. When examining the MGCFA models, different methods of assessing goodness-of-fit is applied. First, the raw chi-squared value for the model of interest is reported. However, it has been shown that inferences based on the raw chi-squared values tend to favour problematic models when the sample size is small and reject sensible models when the sample size is large (Hooper et. al, 2008). In this case, where the sample size is almost 30,000, it is expected that an increase in model complexity will lead to a significant increase in chi-squared value. Second, the Comparative Fit Index (CFI) is reported, where values above 0.95 is considered ‘good model fit’ (Hu and Bentler, 1999). When comparing competing models, Chen (2007) suggests that a decrease in CFI of less than 0.01 is acceptable. Finally, the two absolute fit measures, namely the Root Mean Square Error of Approximation (RMSEA) and Standardized Root Mean Square Residual (SRMR) are used for model comparison.