50 Factors associated with differences in digitally measured student perceptions of teaching quality Table 3.1 Overview of the variables included in the analyses Level Variable Response format Score Student-level Gender Boy/girl Dummy coded (0/1) Level of academic performance Average mathematics grade per student at the beginning of the research period (provided by the teacher, range from 1 to 10 according to the Dutch grading system) Dummy coded (0/1/2) Low-performing group: 1.0 to 5.9; Middle-performing group: 6.0 to 7.4; High-performing group: 7.5 to 10.0 Teacher-level Gender Male/female Dummy coded (0/1) Age In years Years Likability “I like my teacher very much”, one item in the paper-based questionnaire using a 4-point Likert scale Average class score Items scored on a 4-point Likert scale Teaching experience In years Years Initial teaching quality Responses to reformulated items of the Impact! Questionnaire (about the quality of teachers’ teaching in general), using a 4-point Likert scale Average class score Items scored on a 4-point Likert scale Class-level Timing Before noon/after noon Dummy coded (0/1) Class size Number of students in class Number of students Average grade The teacher was asked for the students’ mathematics grades. Average mathematics grade for the class at the beginning of the research period Ethnic diversity The country where students were born (answer options were The Netherlands, Turkey, Suriname, Morocco, Indonesia, Serbia, Germany, Poland, other) Number of different ethnicities represented in the class. Competencies (PIAAC; see Rutkowski et al., 2014). In the following sections, the two models are described. 3.3.6 The item response theory model Item response theory (IRT; Lord 1980) was used to model the students’ responses to the Impact! questionnaire. All items (k = 15) are associated with a unidimensional latent variable that is theorized to be the quality of a teacher’s teaching. The model used is what is called the generalized partial credit model (GPCM, Muraki, 1992). A response to item k in category m (m = 0, 1, or 2) pertaining to a teacher j, by a student i, at time point t, is denoted by Item response theory (IRT; Lord 1980) was used to model the students’ responses to the Impact! questionnaire. All items (k = 15) are associated with a unidimensional latent variable that is theorized to be the quality of a teacher’s teaching. The model used is what is called the generalized partial credit model (GPCM, Muraki, 1992). A respons to item k in category m (m = 0, 1, or 2) pertaining to a teacher j, by a s udent , at time point t, is denoted !"#$ = . The probability of the response is given by: ( !"#$ = | !"#) = %&'()*!+"#$,-!%) /0∑ %&'(2*!+"#$,-!&) & , (1)
RkJQdWJsaXNoZXIy MjY0ODMw