We base our SEM analysis on maximum likelihood (ML)
estimation. Although this method requires multivariate
normality of data (Byrne, 2009), several simulation studies
(e.g., Lei and Lomax, 2005; Boomsma and Hoogland,
2001) have shown that it is apparently quite robust against
the violation of the normality assumption, leading to only
marginally biased parameter estimates. However, standard
errors may be underestimated, thus leading to spurious
results regarding the statistical significance of regression
weights (Byrne, 2009). Therefore, we additionally perform