Teacher’s attitudes: Revising factor structure and construct validity of the interaction with people with disabilities scale

: Teachers’ positive interactions with people with disabilities improve their awareness and advocacy for inclusion, which is an important policy and community aim worldwide. Therefore, promote inclusive efforts in Angola as well as advances in research about measurement of attitudes we conducted a study to analyse the factor structure, validity, and reliability of the Interaction with People with Disability Scale (IDPS) for a sample of 1352 elementary teachers. Results of an exploratory factorial analysis showed a solution with factors of discomfort, coping, and vulnerability explaining 64.29% of the variance. A confirmatory factorial analysis suggested a model with good fit and convergent validity for discomfort and coping, and good evidence of discriminate validity for the three factors. We discussed how exploring attitudes towards people with disabilities with the IDPS can be useful when promoting activities for teachers training in inclusive education.


Introduction
The government of The Republic of Angola has a policy of promoting inclusion, with the aim of valuing individual differences and empowering individuals.This policy agenda has been developed through endorsement of international conventions, strategies, and commitments, by launching a national legal framework, and increasing the involvement and contribution of civil society.
Within the scope of these public policies, since 2012 the People with Disability Act aims to establish a global and integrated policy of prevention, qualification, rehabilitation, and full participation of people with disabilities in social life.To promote full participation in the community, the adaptation of jobs is also recommended, and the executive has set up employment guarantees (Lei [Law] No. 21/12, 2012).Also, in 2016 the Accessibility Act made substantial progress in eliminating barriers and building a global, coherent, and ordered accessibility system that provides people with disabilities or mobility impairment with guarantees of citizenship, and full participation and access to buildings, parks, roads, transportation and indoor and outdoor facilities (Lei [Law] n. º 10/16, 2016).
Finally, at the educational level, a national special education policy for inclusion was adopted from 2017 onwards, based on philosophical and conceptual principals from national legal documents, as well as from international standards set out by the UNESCO and United Nations (Decreto Presidencial No. [Presidential Decree No.] 187/17], 2017).António et al. (2021) analysed the course of the Angolan special education policy from the year of its implementation in 1979 and considered that it is both a specific and separate field in the education system and is based on a medical model of disability.Additionally, also evident in the political discourse is the intention to make the educational system more inclusive, despite the fact that there is still a long way to go (António et al., 2021).
In the above-mentioned legal context, there is a visible commitment to guaranteeing an inclusive society.Then, considering that teachers are citizens who play an important role in making young children, parents, and the civil sectors aware of, and active in advocacy, for an inclusive society, it would be important to understand the attitudes of this community when around someone with such significant challenges.Attitudes toward people with disabilities have a long and complex history (Kauffman & Hornby, 2020) and, as suggested by Sánchez et al. (2021), positive or negative attitudes have a direct influence on societal behaviour.In fact, attitudinal barriers may exclude people with disabilities from having equal opportunities in employment, education, social services, lifestyle, and overall quality of life (Brown et al., 2009;Myong et al., 2021) when interacting with people with disabilities on the personal level (Wallymahmed et al., 2007).It was developed in Australia by the Community Disability and Ageing Program (Gething, 1993), and released in 1991 by Getting (Forlin et al., 1999), to overcome problems associated with administration and psychometric properties of other instruments, as well as to measures attitudes at

3/12
a personal level (Gething, 1993).As underlined by Gething (1993, p. 291), it took in consideration that there is a contrast between the person's reactions "about the way they think people with disabilities should be treated by society (societal level) and their own personal reactions to interaction with people with disabilities (personal level)".The scale explores the motivations and emotions considered to underlie negativism (Gething, 1993), with a focus on the theoretical position that discomfort stemmed from a lack of familiarity or information, and uncertainty (Bania et al., 2020), or from general fear (Thomas et al., 2003).In summary, the IDPS items tap personal reactions related to a) fear of the unknown associated with being unsure of how to behave, or what to expect, when in front of a person with a disability; b) threat to security or to the view of the world as fair when in front of a perceived vulnerability of other or of oneself; and c) guilt, and aversion to difficulties and vulnerability (Gething, 1993).The original IDPS consists of 20 items reflecting six factors named discomfort (items 9, 11, 12, 16, 17, and 18), coping/ succumbing (items 1, 2, 3, and 13), perceived level of information (items 3, 6, 9, 10, and 12), vulnerability (items 7, and 20), coping (items 14, and 15), and a second vulnerability factor (items 4, and 5).
Due to lack of consistent loading, items 8 and 19 were dropped from the initial version of the scale, producing a final one with 18 items (Gething, 1994).The internal consistency of the scale was assessed and it was obtained a Cronbach´s alfa coefficient between .74 and .86(Gething, 1994).
The psychometric properties of the IDP scale have been investigated in several societal contexts, but most of them failing to replicate the original 6-factor structure (see examples in Table 1).It must be underline that the only consistent factor has been the one labelled Discomfort.Therefore, after taking into consideration the aforementioned Angolan context as well as the need for a valid instrument to measure attitudes toward people with disabilities in the country, the aims of the current study were twofold.The first one was to identify the factor structure of the IDPS for Angolan teachers, and the second aim was to determine the model that offers the best fit to the data of the IDPS for Angolan teachers, as well as its convergent and discriminant validity, and reliability.

Method Sample
This study included 1352 teachers from 197 elementary schools in one large province in southwestern Angola (see Table 2), with an average age of 37.58 (SD = 6.94) years.Participants were split into two subsamples using a random sampling of 50% of all cases.One subsample was used for Exploratory Factor Analysis (EFA sample M Age = 37.61, SD = 6.70), and the other one was used for Confirmatory Factor Analysis (CFA sample M age = 37.56, SD = 7.17).

Data Collection Instruments
It was used both a modified version of the 18 items of the IDPS, as designated and reported by Gething (1994) and Forlin et al. (1999), and a sociodemographic questionnaire.We started with a forward and backward procedure for translation of the IDPS into Angolan Portuguese.Then, two researchers met to discuss any issues in this translation, specifically related to the use of 'people-first language' and scientific terminology related to disabilities that was in use and well understood in the country.Finally, items were reviewed in a joint session with four Angolan teachers, who had extensive teaching experience and in-depth knowledge of the characteristics of the population from which the sample was selected.After that we conducted a pilot-test with a first version of the IDPS.We asked 50 regular-education teachers in Angola to fill in the IDPS and to write down comments related to the comprehension and terminology of each item.As a result of this pilot study, Item 16 could not be understood and was dropped, and item 14 was written in a positive way and not negative as in the original.Due to cultural issues, we decided that a 4-point scale ranging from 1 (strongly disagree) to 4 (strongly agree) would be used, with a higher score indicating a greater level of uneasiness in social interaction with people with disabilities.
Additionally, we renamed the scale."Interaction with People with a Disability Scale," reflecting a 'people-first language' approach as in Forlin et al. (2001).In addition, each participant completed sociodemographic questions on age, gender, years as a teacher, professional qualifications, and level of contact with people with disabilities.

Procedures
After ethical approval from the scientific com-

Data Analysis
During data screening and preliminary analysis of the large sample, normality was tested with skewness and kurtosis values and "either an absolute skewness value ≤ 2 or an absolute kurtosis (excess) ≤ 4 was used as reference value for determining considerable normality" (Mishra et al., 2019, p. 70)

Descriptive Statistics
Table 3 presents the descriptive statistics for the answers of the teachers.We underline that on the one hand, variable 15 presents a value of one for the mode, and as this is a reversed variable, it shows that most teachers do agree and strongly agree that after frequent contact, they notice the person, not the disability.On the other hand, variable 1 presents a value of four for the mode, showing that most teachers agree that it is rewarding when they are able to help a person with a disability.Moreover, all distributions show acceptable skewness and kurtosis.Distributions of items 1, 2, 3, 6, and 13 present a negative sign for skewness, indicating fewer participants are easy then uneasy in social interaction, while the rest of the items present a positive sign, meaning that fewer participants are uneasy than at ease in social interaction.

Exploratory Factor Analysis
With the EFA sample we conducted an initial exploratory factor analysis using principal

Confirmatory Factor Analysis
The four-factor model fit was evaluated using confirmatory factor analysis on data of the CFA sample (see Figure 1).

Convergent and Discriminate Validity
Results on testing for convergent validity showed that all items except item 3 (coping) presented standardized loading estimates higher than .5 (λ ij ≥ .5).Item 3 presented a factor loading of

9/12
.49 and so more of the variance is error variance and not explained variance.The model presented an AVE of less than .5 in each of the three factors that is an indicator of lack of convergent validity.The last indicator of convergent validity is reliability.Cronbach's alpha and CR presented good values for discomfort and coping but not for vulnerability.Following recommendations by Gaskin et al. (2019), items 3 and 18 were removed to accomplish convergent validity for discomfort and coping.Finally, all factors revealed good evidence of discriminate validity, when considering that the AVE value for any two constructs is greater than the inter-construct squared correlation estimates between these two constructs (values above the diagonal in Table 4).This is also observed when we used the test of squared root of AVE being greater than the correlation estimates between these two constructs or MSV being less than the AVE (see Table 5).

Discussion
The findings of this study extended international scientific knowledge of the factor structure of the IDPS for the Portuguese language and for a sample of regular teaches in Angola.Results of the exploratory factor analysis suggested that the data from the sample EFA was best represented by a 10-item structure within three factors, unlike the 18 items and the six factor structure of the original scale.Factor 1 (discomfort) consisted of items 9, 11, 12, 17, and 18.Generally, it captures the degree to which interacting with a person with a disability results in uncomfortable, unsure, afraid, and quick contact or the fact that they cannot help staring at them.Discomfort was considered as the main and the most stable factor, which is in line with the findings of previous English versions of the scale (Forlin et al., 1999;Iacono et al., 2009;Maclean & Gannon, 1995;Tait & Purdie, 2000;Thomas et al., 2003;Wallymahmed et al., 2007), as well as Spanish (Lobato et al., 2021), Greek (Bania et al., 2020), and Urdu versions of it (Yoshida et al., 2003).Factor 2 (coping) consisted of items 1, 2, and 3, and captures the level to which the participants feel rewarded when able to help, frustrated when they do not know how to help, and hurt when see that people with disabilities want to do something and cannot do it.
Finally, factor 3 (vulnerability), contained items 5 and 7, and captures the perception of own vulnerability, related to the feelings of a potential disability, and the gratefulness for not having it.
Similarly, it replicated the study conducted by Thomas et al. (2003), who found a three factors model within 14 items when analysing attitudes of undergraduate students from the USA towards individuals with disabilities.The items included in the three factors in our sample were included in the same factors in Thomas et al. (2003).Furthermore, results of the confirmatory factor analysis advanced our knowledge related to the model that offers the best fit to the data of the IDPS for Angolan teachers, as well as its convergent and discriminant validity and reliability.The measurement of the IDPS three-factor model provided a good fit, and convergent validity for discomfort 10/12 Educação, Porto Alegre, v. 47, n. 1, p. 1-12, jan-dez. 2024| e-45156   and coping after elimination of items 18 and 3.
There is a lack of convergent validity for coping, which means that for this factor "on average more error remains in the items than variance held in common with the latent factor upon which they load" (Hair et al., 2019, p. 676).Although reliability values were good for discomfort and acceptable for coping and vulnerability, we can argue that Cronbach's alpha is influenced by the number of items (Field, 2013) and these factors have four, and two items, respectively.All factors presented good evidence of discriminate validity, showing that these constructs are truly distinct from each other, capturing some phenomena others do not (Hair et al., 2019).These findings suggest that major developmental work is required in our model.First, we need to change the wording of item 16 in order to include it in the analysis, and on the fact that items 6, 13, 14, and 20 failed to load on any of the three factors due to small communalities and that item 15 presented problematic cross loadings.
Secondly, in order to improve convergent validity in discomfort and coping items 18 and 3 were removed, and to improve composite reliability on vulnerability, item 7 needs to be removed.In this way, as the grouping of the factors has resulted in two factors that contain only two items each, additional items should be changed or further developed, so factors are represented by at least five items each.In summary the final IDP has the following factor structure: It consists of eight items reflecting three factors named discomfort (items 9, 11, 12, 16, and 17), coping (items 1, and 2), and vulnerability (items 5, and 7).Thus, adjustment and further evaluation of some items need to be considered to improve psychometric soundness for future use.These results corroborated previous conclusions that the IDP requires refinement in terms of its items and that the factor structure is unstable (Bania et al., 2020;Forlin et al., 1999, Iacono et al., 2009;Loo, 2001;Wallymahmed et al., 2007).Despite this need IDP does provide a measure of attitudes towards disability that are consistently positively associated with education or experience with disability (Wallymahmed et al., 2007), which are two characteristics of our sample.
This study has key implications for inclusion, in a way that, as suggested by Loo (2001), the IDPS can be especially useful for purposes other than decision-making, as such use does not rely on strong psychometric properties of the measure, although that is desired.This means using the IDPS to stimulate discussion, critical reflection, and self-awareness within organized informational sessions in municipal members' meetings, civil associations that advocate for people with disabilities, and within teacher pre-service and in-service training and support.This is particularly important for today's teacher training for inclusion provided by the National Institute of Special Education, as well as for the training that is beginning to be undertaken in the Angola university system with the introduction of curricular units related to a successfully inclusive education for students who have disabilities, are at risk or with gifted skills, and their families.A measure of uneasiness during interactions with people with disabilities provides a starting point or a direction towards enriching areas of greatest concern (Forlin et al., 2001), as well as revealing how perceptions change in response to a specific training (Forlin et al., 1999;Iacono et al., 2009;Tait & Purdie, 2000).

Limitations and Future Directions
We identified three limitations in the scope of this research.First, the nature of the sample limits generalising the results to the country.Therefore, we recommend a randomly selected sample in-

11/12
It would be an asset to continue to improve this measurement as well as developing others that are not self-reported and reflect behaviours and actions, as suggested by Thomas et al. (2014).
Considering the commitments made at national and international levels throughout the years, public policies have sought to create the conditions for implementing the civil, political, educational, economic, social, and cultural rights of people with disabilities enshrined in the Angolan legal system and in international human rights legal instruments.In particular, this research has generated findings that have practical implications for assisting in facilitating preservice and in-service teachers' training, and public awareness campaigns, promoting high quality in services, positive attitudes, accessibility, and the latest scientific evidence for the benefits of people with and without disabilities in the Republic of Angola.
. An appropriate measurement of this attitudes would support the use of research-based information to raise awareness throughout society, and specifically in connection with the country's incipient teacher training efforts.In fact, measuring public and professional attitudes to persons with disabilities has received much attention over the last four decades all over the world, being The Interaction with Disabled People Scale (IDPS) the most widely used and validated scale to measure levels of discomfort that an individual might feel mittee of one public university in Portugal to carry out the study, we obtained authorizations from representatives of the Angolan provincial governments, directorates, and participating teachers.The second author distributed and collected copies of the IDPS and of the demographic questionnaire in 197 regular schools.Each participant's consent was obtained after a presentation about the procedure involved with the data collection as well as with ethics.Participants completed the questionnaire and the demographic information individually, under neutral, untimed conditions and with no researchers present, with safeguarded confidentiality and anonymity.The return rate was about 93%.
. Missing data were randomly existent and with exceptionally low percentage in extension, and those cases were deleted, as were outliers with z scores ≥ 4 identified with boxplots and case summary, as suggested byHair et al. (2019).Before the analysis, a reverse-coding transformation was applied to items 10 and 15.An exploratory factor analysis (EFA) using principal component analysis with orthogonal varimax rotation was conducted on a sample with a ratio of about 38 participants to 1 variable to be analysed.For EFA statistical adequacy we used an anti-imaging correlation matrix showing small partial or anti-image correlations, communalities of .50 and higher, a Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy for the overall test and for individual items greater than .50,and a significant Bartlett's Test of Sphericity (p < 0.05)(Hair et al., 2019).A confirmatory factor analysis (CFA) was a way of testing how well the pattern of factor loadings of the measured variables represents the latent constructs of the obtained model in EFA.Considering the sample size and the number of observed variables, besides the overall chi-square goodness-of-fit test (χ²) and the p value associated with it, the following Fit Indices and cut-off values were associated with a model that fits well, in line with Hair et al. (2019).First, the following absolute fit measures: the chi-square statistic divided by the degrees of freedom (χ²/df) less than 2.0 being considered very good, and between 2.0 and 5.0 acceptable; the Root Mean Squared Error of Approximation (RMSE) less than .07; the 90 percent confidence interval for RMSE and the Goodness-of-Fit Index (GFI) greater than or equal to .95; and the Standardized Root Mean Square Residual (SRMR), with the conservative cut-off value of .05.Second, the following increment fit indices were used: Comparative Fit Index (CFI), greater than or equal to .95, and Tucker-Lewis Index (TLI) greater than or equal to .90.Model modifications were conducted using the modification indexes and standardized residuals as guidelines to develop a better fitting (Hair et al., 2019).The final stage of the CFA involved the analysis of construct validity of the model with both convergent and discriminant validity (Hair et al., 2019).Convergent validity was evaluated in this study using the analysis of the size of factor loading (λ ij), the Average Variance Extracted (AVE) greater than or equal to .50, as well as the reliability indicators Cronbach's Alpha and Composite Reliability (CR) greater than or equal to .70.Discriminant validity was analysed with the Fornell and Larcker (1981) test conside-ring that the AVE value for any two constructs is greater than the squared of the correlation estimates between these two constructs(Hair et al., 2019).Additionally, we looked to both the test of squared root of AVE for the factors being greater than the correlation estimates between these two constructs and MSV being less than the AVE(Gaskin et al., 2019).To perform factor analysis, we randomly split participants into two data sets (Lorenzo-Seva, 2021), with approximately 50% of all cases via SPSS with a seed set at 2000000.One dataset was used for exploratory factor analysis and the other for confirmatory factor analysis.All statistical analyses were undertaken in version 27.0 Statistical Package for Social Sciences, SEM AMOS, as well as in Gaskin et al. (2019) Master Validity Tool and AMOS Plugin.
component analysis with orthogonal varimax rotation on the 17 items.Visual inspection of the anti-imaging correlation matrix shows small partial or anti-image correlations, and measure of sampling adequacy values of each individual variable in the main diagonal exceeding .69which is considered middling (Hair et al., 2019).The overall Kaiser-Meyer-Olkin test was .842which is meritorious (Hair et al., 2019), and the Bartlett's Test of Sphericity was significant (χ 2 (136) = 2934.75,p < 0.001), indicating that sufficient correlations exist among the variables to proceed (Hair et al., 2019).Several recalculations resulted in the remo-val of items 6, 13, 14 and 20 due to communalities lower than 0.50, and item 15 due to problematic cross loading (Hair et al., 2019) evaluated by the 1.2 ratio of the squared loadings.Therefore, recalculating all the analysis a final KMO = .782and all-measure of sampling adequacy values for individual items greater than .68 were obtained, as well as significant Bartlett's test of sphericity (χ 2 (45) = 1947.095,p < 0.001).The scree plot shows inflections that justify retaining three factors, which have eigenvalues over Kaiser´s criterion of 1 and in combination variance explained of 64.29%.
of the cases and no significant cross-loading, which is considered indicative of a well-defined structure for the sample.These analyses show that the six-factor model do not fit the Angolan data.

Figure 1 -
Figure 1 -Confirmatory Factor Analysis cluding teachers from the 18 municipalities of the country, and from one.A second limitation of the study is the lack of other instruments that might serve as a gold standard measure for exploring perceptions of people with disabilities that have previously been validated into Angolan Portuguese.Third, lack of experience in participating in research, and of the content of the items in the scale, produced participants' curiosity regarding the study and generated motivation for taking part in it, but may have introduced biased answers due to desirable responding.Two future research implications from this study are derived.

TABLE 1 -
IDP FACTOR CHARACTERISTICS FOR DIFFERENT LANGUAGES AND SOCIETAL CONTEXTS

TABLE 2 -
SOCIODEMOGRAPHIC CHARACTERISTICS OF PARTICIPANTS

TABLE 3 -
STATISTICS OF THE IDPS

TABLE 4 -
FACTOR ANALYSIS OF THE IDPS

Table 4
displays loadings exceeding .73 in most Significance Level: * p < 0.010 ** p < 0.001.The Diagonal elements in bold are squared root of AVE, values above these diagonals are squared correlations, and below are correlations estimates among factors.