Statistical Analysis

First, we needed to determine whether the three scales we used in the study measured the same constructs in both countries, i.e., that they demonstrated measurement invariance across the Polish and Vietnamese samples. Thus, we assessed the three scales’ cross-country equivalence through multigroup confirmatory factor analysis (MGCFA). In the beginning, the factorial structure of each scale was assessed separately for Polish and Vietnamese samples using CFA. To assess the fit of the models, we followed Brown (2015), using the following criteria: CFI > 0.90 and RMSEA < 0.08 (e.g., Brown, 2015). However, Kenny et al. (2015) showed that RMSEA often underestimates fit when the degree of freedom is small, so we used an SRMR criterion <0.08 for the UWES-9S and the Study Satisfaction Scale.

In the steps following, we tested the measurement invariance of the three scales we used in the study in Poland and Vietnam. In cross-country research, we usually estimate three levels of invariance: configural, metric, and scalar. Each of them is defined by the parameters that are constrained to be equal across samples (Milfont and Fisher, 2010; Beaujean, 2014). Configural invariance is present if in all groups the measurement model is built of the same number of factors that consist of the same indicators; metric invariance is determined when factor loadings are equal across the groups; and scalar invariance requires that factor loadings and all intercepts are equal across the groups. It is also possible to determine partial invariance, which is considered to be sufficient for cross-group comparisons (Byrne et al., 1989). Partial invariance requires that the parameters of at least two indicators per construct are equal across the groups.

We started the measurement invariance investigation by testing for configural invariance across the Polish and Vietnamese samples. To identify subsequent levels of measurement invariance (metric and scalar), we used the following cut-off criteria: ΔCFI ≤ 0.01 (see Chen, 2007). The R environment (R Core Team, 2018) supported by the lavaan package (Rosseel, 2012) was used to conduct measurement invariance analysis using maximum likelihood with robust standard errors estimation (MLM).

We next compared the significance of differences between the average results of the studied variables between Polish and Vietnamese students. For this purpose, we used the t test for independence samples. We also conducted a linear regression analysis to test the hypothesis about the moderation role of a country in the relationships between filial piety and study attitudes (engagement and satisfaction). Our model included two independent variables (RFP and AFP): one moderator (country) and two dependent variables (study engagement and study satisfaction) (see Figure 1). Finally, to illustrate the moderation effect, we conducted a linear regression analysis separately on the Polish and Vietnamese samples.