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

of the total variance. Factor loadings ranged from .70 to .80, and communalities ranged from .48 to .63. The scale had high internal reliability (α = .94). Self-efficacy in creating high-quality student-interactions. Based on the data distribu- tions and anti-image correlation matrix, four items were deleted (see Table 1). PCA on the remaining 8 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 5.18, which explained 64.7% of the total variance. Factor loadings ranged from .70 to .85, and com- munalities ranged from .49 to .72. The scale had high internal reliability (α = .92). Step 2: CFA with in-service teachers. In the second step, the solution that resulted from the first step was checked for in-service teachers using CFA. With regards to self-efficacy, we included four different factors: noticing diversity with six items, enabling high-quality stu- dent-interactions with eight items, creating stimulating learning environments with 13 items, and collaborating with colleagues and parents with 16 items. With regard to beliefs about diverse students, we included four factors geared toward specific groups, being ethnic minority students, disabled students, students from low SES, and essentialist notions regarding the gen- der binary and sexuality with respectively six, six, five, and nine items. For beliefs on the responsibility of the educational field to create education for all, we included three factors

focusing on general school policy, and initiatives geared specifically towards either ethnic minority students or students with a disability, with respectively eight, four and four items. The result of this points to acceptable but not good fit indices (AIC = 155650, X2 = 8030, DF = 3430, p < 0.001, X2/DF = 2.34, RMSEA = 0.048, CFI = 0.760, SRMR = 0.066). Changes were made in order to achieve a better fit. As factor loadings for the items related to collaborat- ing with colleagues were consistently lower (i.e., between .424 and .511), we tested a model with a separate factor pertaining to collaborating with diverse parents (consisting of 10 items) and one pertaining to collaborating with colleagues (consisting of six items). These adaptations led to a better fit (AIC = 154671, X2 = 7125, DF = 3419, p < 0.001, X2/DF = 2.08, RMSEA = 0.043, CFI = 0.807, SRMR = 0.059). In a next step, 13 items were deleted as sug- gested by the modification indices, R-square values or factor loadings (see Table 1). This led to a further improved model (AIC = 133214, X2 = 4765, DF = 2418, p < 0.001, X2/DF = 1.97, RMSEA = 0.041, CFI = 0.846, SRMR = 0.054). In accordance with suggestions by Hu and Ben- tler (1999) and as shown by Coubergs and colleagues (2017), we accept a combination of CFI- values approaching .90 with RMSEA values smaller than .06 and SRMR-values under .08 as good model fit. Factor loadings can be seen in Tables 2 and 3 (note that the loadings for self- efficacy and beliefs are shown in separate tables for lay-out reasons only, as these are tested as

one model within CFA). This table shows loadings above |.50| for all items, with the exception of C2_14, C2_27, C2_29, C5_14, and C5_15. As these items still loaded above |.40|, while delet- ing them was not conducive to the Cronbach’s alpha of the respective factor, this was deemed acceptable. All in all, this results in a model with five factors related to self-efficacy and seven factors related to beliefs, with alpha scores being good for each factor and showing that no items could be deleted in order to get a higher alpha. More specifically, the factors related to self-efficacy comprises “noticing student diversity” consisting of 6 items with α = .87, “enabling high-quality student-interactions” consisting of 6 items with α = .85, “creating stim- ulating learning environments” consisting of 13 items with α = .90, “collaborating with col- leagues” consisting of 3 items with α = .83, and “collaborating with parents from diverse backgrounds” consisting of 8 items with α = .88. The factors relating to beliefs comprises pro- fessional beliefs on different types of diversity, specifically on ethnic minorities (# of items = 6, α = .79), students with a disability (# = 4, α = .78), students from low-SES families (# = 5, α = .76), and essentialist notions on sexuality and the gender binary (# = 8, α = .79), as well as beliefs on the responsibility of the educational field to create education for all, including a PLOS ONE Measuring teachers’ competences for inclusive education PLOS ONE | https://doi.org/10.1371/journal.pone.0291033 November 8, 2023 12 / 22