79 4 Figure 4.6 Average teaching quality (a time period includes 5 days). Due to a holiday period in the Netherlands, no evaluations of teaching quality happened during time periods 7, 8 and 9. Time period 18 was a holiday period as well, but because not all schools had off in the same week, two measurements were conducted in that period, which explains the small decline in this period. In the null model, we see that teachers varied considerably in perceived teaching quality. Most of the variance (73.4%) in the measurements was situated at the teacher level (intercept variance). Thus, most of the variance was teacher specific, and only a relatively small part related to the residual variance. In model 1, no significant effect was found for time (in linear terms). In model 2, the quadratic term for time was added, which showed a significant effect of this parameter (p < 0.01). The variance of “time” declined from .026 to .020, so the quadratic trend investigated with this model partly explained the variance. Models 3 and 4 were estimated to investigate the existence of random slopes. Because both models showed no statistically significant effects, there was no evidence for such random slopes, which implies that teachers did not have different growth curves. The results in Table 4.2 show the best fit for model 2, because the deviance statistic was lowest there, and parameters were parsimoniously added in the model. The predicted values for teaching quality in model 2 were plotted against date, which showed a quadratic trend of the predicted teaching quality over time (see Figure 4.7) and therefore supported the results as presented in Figure 4.6.
RkJQdWJsaXNoZXIy MjY0ODMw