The longitudinal association between children's growth mindset in senior primary school and their parents' growth mindset · Chen J, Liu C

Results

Table 1 shows the mean, standard deviation, and correlation coefficient of the primary variables. According to the results in Table 1, the correlation for children’s growth mindset at the five time points was between 0.243 and 0.492; for fathers’ growth mindset, it was between 0.195 and 0.334; and for mothers’ growth mindset, it was between 0.208 and 0.467, which indicates that their growth mindset had a moderate degree of stability. In addition, the children’s growth mindset scores were positively correlated with the fathers’ scores (r’s = 0.056–0.195) and mothers’ scores (r’s = 0.057–0.179). In general, the parent–child correlation was smaller than the intra-individual correlation.

Table. Descriptive statistics and Pearson correlations for the main study variables. (See the original article for the full table.)

Table. Longitudinal confirmatory factor analysis on growth mindset. (See the original article for the full table.)

A longitudinal measurement invariance analysis for children’s growth mindset showed that the configuration invariance model fit well, and the weak invariance model also fit well. Comparing these two models, Δχ2 = 17.191, Δdf = 12, p > 0.05, ΔCFI = 0 < 0.01, ΔTLI = 0.003 < 0.01, the weak invariance model was accepted. Then, comparing the weak invariance model and strong invariance model, Δχ2 = 40.053, Δdf = 12, p < 0.05, ΔCFI = 0.001 < 0.01, ΔTLI = 0.001 < 0.01, we showed that strong measurement invariance was satisfied. The results suggested that the structure of children’s growth mindset remained invariant of the factor loadings and intercepts across the five time points. The measurement invariance of fathers’ growth mindset and mothers’ growth mindset was similar to that of children’s (for specific parameters, refer to Table 2).

Univariate growth trajectories

The model fit for the univariate growth trajectory of growth mindset was excellent for children/fathers/mothers, and the parameter estimates are shown in Table 3. Figure 2 shows the developmental trajectory of growth mindset. In the unconditional latent growth curve model for the children, the initial level of the growth mindset was 3.431 (p < 0.001); that is, the mean of the children’s growth mindset was 3.431 at the fifth measurement and showed a linear downward trend over time at the five time points of the test, with a slope of −0.011 (p = 0.008). The correlation coefficient between the intercept and the slope was r = −0.015, p < 0.001, which indicated that the children’s growth mindset at T5 was inversely related to the rates of change over time. In addition, both the intercept variance (σ2 = 0.262, p < 0.001) and the slope variance (σ2 = 0.021, p < 0.01) were significant, which indicated significant individual differences in both the initial level and the change in the growth mindset of senior primary school children. The parental unconditional latent growth curve models were similar to the children’s model, and the slope factor of growth mindset was significant and negative (b = −0.019, p < 0.01, for fathers, b = −0.016, p < 0.01, for mothers). Significant intercept and slope variance indicated significant individual differences in starting levels and rates of change in fathers’ and mothers’ growth mindset. However, there was a statistically significant and negative correlation between the intercept and slope in the mothers’ model but not in the fathers’ model.

Table. Parameter estimates and model fit for univariate growth models. (See the original article for the full table.)

Figure. The developmental trajectory of growth mindset. GM, growth mindset. (See the original article for the figure.)

Parallel process latent growth curve model

Then, a parallel process latent growth curve model was established that included the trajectory of children’s, fathers’, and mothers’ growth mindset. First, the three univariate growth curve models mentioned earlier were combined within one parallel process latent growth curve model. This unconditional model indicated an adequate fit, χ2(78) = 225.761, p < 0.001, CFI = 0.986, TLI = 0.981, RSMEA = 0.023. Then, a conditional parallel process latent growth curve model was established that controlled for children’s gender and age, father’s education level, mother’s education level, and annual family income. This conditional model also fits the data well, χ2(143) = 290.002, p < 0.001, CFI = 0.986, TLI = 0.982, RSMEA = 0.017. The parameter estimates are reported in Table 4. In the conditional model, children’s gender was positively associated with the intercept of children’s growth mindset (β = −0.008, p < 0.01), while there was no significant correlation among children’s age, father’s education level, mother’s education level, annual family income, and the intercept/slope of children’s growth mindset.

Table. Parameter estimates for parallel process latent growth curve model. (See the original article for the full table.)

An unconditional model showed that the initial level of maternal growth mindset could positively predict the initial level of children’s growth mindset (β = 0.127, p = 0.03), which means that maternal growth mindset at Time 1 positively predicts children’s growth mindset at Time 5. However, this path was marginally significant after controlling for demographic variables (β = 0.109, p = 0 0.064) in the conditional model. Moreover, the paths from the slope of maternal growth mindset to children’s intercept were both significant in the conditional (β = 0.215, p = 0.008) and unconditional models (β = 0.227, p = 0.007), which indicates that the change in mothers’ growth mindset can influence the level of children’s growth mindset in the fifth test. The results also showed a significant and positive link from the slope of maternal growth mindset to children’s slope in the conditional (β = 0.362, p = 0.006) and unconditional models (β = 0.369, p = 0.007); that is, when the mothers’ growth mindset declined more rapidly, the children’s growth mindset declined more rapidly. In contrast, neither the intercept nor the slope of fathers’ growth mindset predicted the change in children’s growth mindset or children’s growth mindset at Time 5.

Multiple group analyses

Finally, we tested whether any of the paths from the latent growth factors of fathers’ or mothers’ growth mindset to the latent growth factors of children’s growth mindset differed by children’s gender. Eight paths were checked by the Wald test (refer to Table 4). The results showed that only one path test was significant, that is, the path from the intercept of fathers’ growth mindset to the intercept of children’s growth mindset differed by children’s gender [χ2(1) = 4.154, p < 0.042]. Specifically, paternal growth mindset in the first test predicted girls’ growth mindset in the fifth test (β = 0.253, p = 0.014) but not boys’ growth mindset (β = −0.038, p = 0.709).