Structural modeling of relationships among the variables
The initial analysis used SEM to construct the baseline model, M0. This model used closeness TCR and conflict TCR to predict SR, SE, mathematics, and vocabulary. The model fit indices for M0 were excellent, with x²/df = 0/0, CFI = 1, TLI = 1, SRMR = 0, and RMSEA = 0. However, it was noted that the coefficients of the four paths from closeness TCR and conflict TCR to mathematics and vocabulary did not reach statistical significance. To simplify the model, these non-significant paths were removed.
Model M1 was then developed based on M0, which excluded the direct predictive paths of closeness TCR and conflict TCR to mathematics and vocabulary. The purpose of M1 was to examine whether children of different genders and grades exhibited the same model. The results indicated that this model was acceptable and suggested structural constancy among different groups in the path model. All the indicators of M1 met the fit criteria, with x²/df = 5.46/4, CFI = 0.997, TLI = 0.989, SRMR = 0.037, and RMSEA = 0.015. In M1, all path coefficients reached the significance level, with the path coefficient for conflict TCR being negative, while the remaining path coefficients were positive (see Table 4).
<Insert Table 4 HERE>