74 Does smartphone-assisted student feedback affect the quality of teachers’ teaching? group (hypothesis three) and the results were presented. Scatterplots were used to examine the relation between teachers’ initial teaching quality scores and the improvement-oriented actions teachers undertook. Only teachers who completed the digital questionnaires at M1, M2 and M3 were included in this analysis (N = 20). Teachers’ improvement-oriented actions undertaken outside the lesson were also examined by means of bar charts and presented in the results section. To test hypothesis four, graphs of teachers’ teaching quality scores over time were studied. The graphed scores are averages of scores for the quality of their teachers’ teaching as reported by students over a 5-day period. In that way, the scores for one working week were compressed in order to reduce data and to enable a better examination of the trends. Scores on the students’ questionnaires about the quality of their teachers’ teaching at M0 (pretest) and M3 (posttest) were included, as well as all scores from the Impact! tool. The number of measurements per time period of five days was, on average, 7 (SD = 3.92). A multilevel growth model (Field, 2013; Snijders & Bosker, 1999) was used, as it takes into account the nested structure of the data (different time points were nested within one teacher; Peugh, 2010). Consistent with previous research (Leatherdale et al., 2005; van der Scheer & Visscher, 2017), a four-step modelling procedure was conducted as a follow-up to the null model, using SPSS Statistics version 24. Scores for teaching quality at M3 were used as the dependent variables. Time was used as the independent variable (centred at the f irst measurement, 22 November 2016). Values express the number of days that had elapsed since the f irst measurement. Linear and quadratic terms for time were included in the models. In the null model, whether teachers vary in teaching quality scores at M3 was determined. Model 1 was developed with only the linear effect of time as a predictor. In model 2, the linear as well as the quadratic effect of time were added as predictors, and predicted scores were saved to be plotted against the date, and to illustrate the quadratic trend. In models 3 and 4, the strength of the direct effects of both the teacher- and the date-level predictors was computed using random coeff icient regression models (to investigate whether teachers have different slopes). Model 3 only included the time variable (centred around the f irst measurement) as a predictor. For model 4, the time variable was centred quadratically and added as well.
RkJQdWJsaXNoZXIy MjY0ODMw