Education for all in action: Measuring teachers’ competences for inclusive education

disability, with factor loadings ranging from .57 to .72, and a Cronbach’s alpha of .77. The final factor consisted of 5 items tapping beliefs towards low SES-students, with factor loadings rang- ing from .42 to .72, and a Cronbach’s alpha of .72. Responsibility of the educational field to create Education for All. Based on the data distributions, correlation matrix, and anti-image correlation matrix, fifteen items were deleted (see Table 1). PCA on the remaining 18 items revealed satisfactory KMO-value, anti-image correlations and Bartletts’ test, and that it was reasonable to retain four factors. Communalities ranged from .41 to .72. These four factors explained 53.98% of the total variance and had eigenvalues of 25.42, 11.55, 9.04, and 7.97 respectively. The factor correlation matrix shows that there was a correlation between the various factors (r-values between |.134| and |.281|). Interestingly, the revealed factors did not line up directly with the diversity groups included in the items. Rather, the first factor, consisting of 6 items, seemed to tap a broad sense of school responsibility, including general items and items with regards to tackling discrimination based on SES, gender and sexuality. This factor had loadings ranging from .62 to .75, and a Cron- bach’s alpha of .71. The second factor tapped beliefs on school approaches specifically directed at ethnic minority students, consisting of 4 items with loadings ranging from .45 to .86, and a Cronbach’s alpha of .70. The third factor consisted of 4 items representing beliefs on the broad

role teachers should take on, with factor loadings ranging from .59 to .71, and had a Cron- bach’s alpha of .61. The final factor consisted of 4 items tapping beliefs on schools’ responsibili- ties towards students with a disability, with factor loadings ranging from .64 to .83, and had a Cronbach’s alpha of .75. In acknowledgement of this rather surprising factorial structure, we decided for the second step to add a few more items representing a general sense of schools’ responsibilities towards all diverse students, rather than geared towards specific groups. However, the items pertaining to the factor representing teachers’ broad role were dropped from the questionnaire for the second step, due to its low Cronbach’s alpha. Responsibility beliefs towards diverse parents and colleagues. Inspection of the data distributions revealed that respondents had answered the items in a rather extreme manner. That is, the median-likert category was at least 5 = agree for all items, indicating that 50% of respondents scored 5 or higher on scales going from 0 to 6. As this indicated severe ceiling- effects, it was decided to rework the dimension as a whole before submitting the questionnaire to in-service teachers in step 2. It was hypothesized that the stem “It is my job to. . .” was lead- ing to a certain social desirability bias rather than providing a reflection of actual intentions. Consequently, it was deemed advisable to rework the scale to tap respondents’ self-efficacy in dealing with diverse parents and colleagues rather than their responsibility beliefs. Hence, in-

service teachers were presented with 16 items preceded by the stem “To what extent are you able to. . .” and a 9-point Likert scale (0 = not at all, 8 = a lot). Self-efficacy in noticing student diversity. The 9 items had a KMO value of .902 and Bartletts’ test was significant (X2 = 4584, df = 36, p < .001). Inspecting the anti-image correla- tion matrix led to the deletion of three items (see Table 1). PCA on the remaining 6 items revealed satisfactory KMO-value, anti-image correlations and Bartletts’ test. The scree plot and Kaiser criterion indicated one factor to extract with an eigenvalue 4.00, which explained 67% of the total variance. Factor loadings ranged from .76 to .84, and communalities ranged from .58 to .72. The scale had high internal reliability (α = .90). Self-efficacy in creating stimulating learning environments for diverse students. Inspection of the correlation matrix and anti-image correlation matrix revealed several prob- lematic item-sets, so eight items were deleted (see Table 1). PCA on the remaining 13 items revealed satisfactory KMO-value, anti-image correlations and Bartletts’ test. The scree plot and Kaiser criterion indicated one factor to extract with an eigenvalue 7.43, which explained 57.2% PLOS ONE Measuring teachers’ competences for inclusive education PLOS ONE | https://doi.org/10.1371/journal.pone.0291033 November 8, 2023 11 / 22