Changes in perceptions of teaching quality in secondary schools in Rwanda during the COVID-19 global pandemic and the subsequent closing and reopening of schools

The stability of measures of teaching quality is essential for making generalizations of results stemming from these measures to other teaching situations. However, no research has examined the effects of unexpected situational factors on the stability of these measures. Therefore, the purpose of this two-phase quantitative research study was to examine the following aspects among secondary school teachers in Rwanda, using a score-validated, multiple-dimension measure: (a) perceptions of teaching quality (PTQ) prior to the onset of the COVID-19 context (Phase 1; descriptive and correlational design); and (b) the extent to which COVID-19 and the subsequent closing and reopening of secondary schools affected PTQ among STEM teachers in Rwanda, and the associations between these changes in PTQ and selected socio-demographic/locational variables (Phase 2; descriptive and correlational research design). Phase 1 findings revealed that two measures of cultural values (i.e., Attitudes Towards Cultural Values Scale, Inculcating Cultural Values Scale, respectively) generated the most positive attitudes, whereas the Satisfaction with Resources and Material Subscale yielded the least positive attitudes. Phase 2 findings revealed that for four of the nine PTQ scales/subscales, the COVID-19 context negatively affected PTQ. These findings provide compelling evidence of the importance of monitoring PTQ, especially during times of crises. Moreover, these findings have implications for Rwandan educational policymakers, Rwandan administrators, teacher training administrators, and, above all, the teachers themselves, as they all seek to maximize teaching quality in Rwandan secondary schools.

The stability of measures of teaching quality is essential for making generalizations of results stemming from these measures to other teaching situations. However, no research has examined the effects of unexpected situational factors on the stability of these measures. Therefore, the purpose of this two-phase quantitative research study was to examine the following aspects among secondary school teachers in Rwanda, using a scorevalidated, multiple-dimension measure: (a) perceptions of teaching quality (PTQ) prior to the onset of the COVID-19 context (Phase 1; descriptive and correlational design); and (b) the extent to which COVID-19 and the subsequent closing and reopening of secondary schools affected PTQ among STEM teachers in Rwanda, and the associations between these changes in PTQ and selected socio-demographic/locational variables (Phase 2; descriptive and correlational research design). Phase 1 findings revealed that two measures of cultural values (i. e., Attitudes Towards Cultural Values Scale, Inculcating Cultural Values Scale, respectively) generated the most positive attitudes, whereas the Satisfaction with Resources and Material Subscale yielded the least positive attitudes. Phase 2 findings revealed that for four of the nine PTQ scales/subscales, the COVID-19 context negatively affected PTQ. These findings provide compelling evidence of the importance of monitoring PTQ, especially during times of crises. Moreover, these findings have implications for Rwandan educational policymakers, Rwandan administrators, teacher training administrators, and, above all, the teachers themselves, as they all seek to maximize teaching quality in Rwandan secondary schools.
In Rwanda, like in many countries worldwide, the school system quickly switched to emergency remote online teaching and learning (ERTL) during the closure of schools due to the COVID-19 global pandemic. ERTL represents an alternative and unplanned method for delivering instruction from a distance because teachers were not able/ allowed to be physically located in a classroom with their students. Many teachers needed support to deliver ERTL; therefore, the Rwanda Basic Education Board (REB) provided this support via online professional development for teachers that focused on digital skills. In so doing, REB's goal was to maintain the teaching quality that had been established teaching face-to-face prior to the onset of the COVID-19 pandemic. That is, the goal of REB was to ensure stability in teaching quality, despite the change in instructional delivery to ERTL. Although this represented an important objective of REB, to our knowledge no measure was put in place to determine whether this goal had been achieved. One approach to addressing this gap in understanding is to assess teachers' perceptions of teaching quality both before and during the COVID-19 pandemic. This can provide an important indicator of the stability of perceptions of teaching quality, particularly given the direct role this stakeholder plays in actualising this construct.
The stability of perceptions of teaching quality is essential in order to justify making generalizations about these perceptions to other teaching situations. Interestingly, Gaertner and Brunner (2018), who examined the effects of situational factors on the stability of students' perceptions of teaching quality, documented that "student perceptions are reliable and at least moderately stable across certain teaching situations" (p. 176). The situations examined were the following three situational factors: time between assessments, subjects taught by teachers, and students' grade levels. Although these situational factors are worthy of investigation, they represent what could be termed as expected situational factors. However, what would be even more compelling would be to investigate the effects of unexpected situational factors on the stability of perceptions of teaching quality. And no situational factor can be as unexpected as the COVID-19 global pandemic and the subsequent closure of schools worldwide.
The school closures in Rwanda begun on 14th March 2020, only 3 days after COVID-19 had been declared a global pandemic by the World Health Organization (World Health Organization [World Health Organization, 2020), which coincided with the first confirmed case of coronavirus in Rwanda. Alongside these school closures, the Rwandan Government implemented an array of educational programs to ensure that the approximately 3.5 million students could continue to learn. Most notably, the Ministry of Education (MINEDUC), in consultation with its agencies and other Ministries, activated an Education Sector COVID 19 Response Plan. This plan, which aligned with the national COVID-19 Preparedness and Response Plan and the Education Sector Strategic Plan (ESSP;2018-2023/2024) of the Republic of Rwanda Ministry of Education (2010), outlined the priorities of the education sector with regard to education quality. More specifically, the Education Sector COVID-19 Response Plan involved the following five objectives: Objective 1: Ensure continuity of learning for all students in Rwanda; Objective 2: Ensure that schools reopen with appropriate services and measures in place and that students re-enter the formal education system; Objective 3: Ensure the health and safety of students, teachers, and other education personnel; Objective 4: Protect and provide for vulnerable populations, including children with disabilities, girls, and children from lower wealth-quintiles; and Objective 5: Prepare the resilience of the education system against future shocks.
The Education Sector COVID-19 Response Plan, which provided the overarching and guiding framework, budget, and monitoring and evaluation framework, led to the Rwanda Education Board (REB) delivering remote learning through online platforms (Objective 1) that comprised the following three broad initiatives: (a) national radio lessons, with several Government agencies providing technical support; (b) audiovisual lessons, primarily via national television broadcasting, wherein lessons given for subjects with a greater need for visual content, such as science experiments, were given higher priority; and (c) e-learning, including the launching of a YouTube channel by REB, called REB eLearning, with content for students. Although Kinyarwanda was the medium of instruction for Nursery and Lower Primary levels, English was the medium of instruction for all other grade levels. Further, REB bolstered its online learning websites to support remote learning for teachers, which included online professional development for both school leaders and teachers that focused on digital skills. A helpline was created for students, parents, and community members to support their queries associated with remote learning, as well as to generate feedback on the effectiveness of the resources. Also, Short Message Service (SMS) devices were used to communicate, via text messages, essential information and reminders for parents on the learning schedules of students. To facilitate students with disability (Objective 4), the digital platforms were enhanced to incorporate accessibility features, which included text-to-speech and digital communication formats, such as the REB scripts being translated into Braille, Sign Language, and close-caption.
As such, the response by REB to the COVID-19 global pandemic was unique. And because the overall goal of REB was to maintain education quality during the period of COVID-19 and the ensuing school closures, and because teaching quality is widely acknowledged as an important avenue for improving the quality of education for students (Goldhaber, 2016), it was deemed important to evaluate the impact of COVID-19 on perceptions of teaching quality in Rwanda. Such information could help educational policymakers (e.g., REB, MINEDUC) in their assessment of the extent to which Objectives 1 and 2 of the Education Sector COVID-19 Response Plan-the most immediate objectives-were met.
Just before the first confirmed case of coronavirus in Rwanda had been reported, the current researchers had been able to collect data on perceptions of teaching quality from 1,820 STEM teachers representing 358 schools from 14 districts within Rwanda, which enabled the researchers to score-validate their instrument. STEM teachers were the population of interest because improving STEM education at the secondary-school level is a key priority of the Government of Rwanda (Education Sector Strategic Plan, 2018/19 to 2023. Via this scorevalidated instrument, we were in a position to examine the effect of the COVID-19 global pandemic and the subsequent closing and reopening of secondary schools-hereafter referred to as the COVID-19 context-on the stability of perceptions of teaching quality within a STEM context. This was the major purpose of the present two-phase quantitative research study, with Phase 1 representing before COVID-19 (i.e., February/March 2020) and Phase 2 representing during COVID-19 and the School Closures (i.e., March 2021). Specifically, the following three research questions were addressed: 1. Which perceptions of teaching quality yielded the most and least positive responses among secondary school teachers in Rwanda prior to the onset of the COVID-19 context? (Phase 1) 2. What is the effect of the COVID-19 context on the perceptions of teaching quality among secondary school teachers in Rwanda? (Phase 2) 3. How does the effect of the COVID-19 context on the perceptions of teaching quality vary as a function of selected socio-demographic/ locational variables? (Phase 2) Therefore, the current inquiry represents the first study to investigate the stability of perceptions of teaching quality in the times of a worldwide crisis-namely, the COVID-19 context that led to ERTL.

Theoretical framework
Despite the fact that quality of teaching has been linked consistently to student learning and, for the most part, is within the power of schools to change (Singh & Sarkar, 2015), it is surprising that there has been a dearth of studies on the topic of teaching quality in Sub-Saharan Africa. In fact, a rapid review by the current authors of Scopus-indexed documents that were published between 1960 and 2022 led to the identification of only 21 articles on the topic of quality teaching using the keywords teaching quality (n = 8), teacher quality (n = 7), and teacher effectiveness (n = 6). Of these works published during this 62-year period, only two studies involving schools in Rwanda, where the current study took place, namely, Iwakuni (2017) and Dorimana et al. (2022). Specifically, Iwakuni (2017) investigated teaching quality among prospective (i.e., preservice) lower secondary school teachers in Rwanda, with the findings revealing meagre consideration in pre-service training of how student-teachers learn. The study conducted by Dorimana et al. (2022) was focused on obtaining an understanding of the types of teacher-student interactions that increase mathematical performance in senior secondary school, with the teachers' use of purposeful questions and feedback being among the most influential factors observed. However, in both studies, the construct of teaching quality was not explored comprehensively-that is, it was only explored via a limited number of dimensions and via certain stakeholder groups. For example, Dorimana et al. (2022) administered the Mathematical Classroom Observation Protocol for Practices (MCOP 2 ; Gleason et al., 2015Gleason et al., , 2017, a 16-item instrument that measures the two dimensions of teacher facilitation and student engagement. Due to the lack of studies in the area of teaching quality in Sub-Saharan Africa, Carter et al. (2021) used elements of grounded theory (Glaser & Strauss, 1967)-including the application of inductive reasoning, the use of theoretical sampling via a pilot phase, the collection of qualitative data, and the employment of theoretical sensitivity-within a multiple case study design (Stake, 2005), to explore perceptions of teaching quality in the Rwandan context. Their study was guided by the following overarching research question: What are stakeholders' perceptions of teaching quality in Rwandan secondary schools? and the following sub-question: What makes a good teacher?
Constant comparison analysis (Glaser, 1965), coupled with classical content analysis (Berelson, 1952), led to an emergent theory that perceptions of teaching quality in the Rwandan context are both multidimensional and multifaceted in nature. Fig. 1 displays the themes and subthemes that illustrate this multidimensional and multifaceted nature of perceptions of teaching quality among secondary school teachers in Rwanda. In particular, it can be seen that the following three themes emerged with respect to qualities that good teachers possess: Qualifications (i.e., what the stakeholders deemed to be the educational accomplishment[s] and expertise level[s] that teachers must possess), Knowledge (i.e., teachers, at the very minimum, having adequate knowledge of the subject that they teach), and Teacher Disposition (i.e., comprising the attributes of punctuality, passion, role model, cultural values, and motivation).
With respect to how teachers enact quality teaching, the following three themes were extracted by the researchers: Classroom Practices (i.e., the extent to which teachers provide equal opportunities for student participation, provide active student engagement, monitor student learning, and use resources and materials), Teacher Preparedness (i.e., involving the quality of lesson planning and time management demonstrated by the teacher), and Understanding the Whole Learner (i.e., involving the teacher listening/paying attention to each student and understanding each student).
With regard to the outcomes towards which quality teachers work, the following two themes emerged: Immediate Goals (i.e., teachers promoting positive student academic performance and inculcating cultural values in their students) and Future Goals (i.e., teachers facilitating students being capable of applying learning, becoming good citizens, and gaining paid employment).
In order to develop a measure of teaching quality, we used these aforementioned themes and subthemes that emerged from the FGDs in the study of Carter et al. (2021) (cf. Fig. 1 Co-operation and Development, 2018]) that have been utilized internationally with secondary school level students and in the context of teaching STEM subjects. Fig. 2 provides an example of how the findings from the FGDs translated into the measure of teaching quality (i.e., Teacher Survey). The ensuing instrument was pilot tested and score-validated, yielding the final version of the Teacher Survey, which will be described in the Method section.

Method
As noted previously, the research study contained two phases. Each of these phases will be described in the following sections.

. Research design and participants
The first phase involved the use of both a descriptive research design and a correlational research design to examine perceptions of teaching quality via the scales that were informed by the FGDs. Participants were 1,820 STEM teachers, who were selected via criterion sampling, from 358 schools comprising 14 of the 30 districts in Rwanda that represent all four provinces (but not the City of Kigali). These 1,820 STEM teachers completed the survey in February and March 2020.
The majority of these teachers were men (73.2%), were younger than 35 years of age (61.6%), and had less than 10 years of teaching experience (69.5%). In terms of their roles, most of the teachers in the sample were regular teachers (69.0%), followed by school subject leaders (23.1%), and then by heads of department (4.7%). The remaining three roles-which all represented a small proportion-were contract teachers (0.8%), school-based mentors (0.8%), and directors of study (0.4%), respectively. Of the 3% of the teachers who reported teaching with a disability, slightly more than one half (i.e., 58.2%) reported having poor vision. The next most commonly reported disability-representing approximately one third (i.e., 34.5%) of the STEM teachers-pertained to locomotor disabilities. The remaining STEM teacher participants reported disability issues related to blindness, hearing impairment, speech impairment, and mental health disability-each of these disabilities being reported by a single teacher (1.8%). None of these teachers reported having multiple disabilities. Slightly more than three quarters of the teacher participants (75.4%) walked to their schools. The next most common travel mode was private bus (7.4%), followed closely by public bus (6.3%). The two remaining travel modes, respectively, were private bicycle (5.5%) and private motor car (3.0%). In terms of the duration of time that it took each teacher to travel to work, nearly three quarters of the teachers (72.0%) took 30 min or less to travel to work, yielding a mean travel time of 28.00 min (SD = 25.64).

Instruments and procedure
The survey, which was administered face-to-face in Phase 1, contained several sections that included Likert-format items. Each of these sections was subjected to a first-order principal components analysis (PCA) to examine the structure of the underlying items (i.e., correlational research design), specifically to ascertain whether the extracted scale was unidimensional (i.e., representing one scale with no subscales) or multidimensional (i.e., representing two or more subscales). Specifically, within each section of the survey, items were included in the same factor (i.e., scale/subscale) if they were all statistically related (i.e., dependent) with each other and, at the same time, were all statistically independent from the remaining items contained in that section. Consequently, within a section of the survey, a single scale was extracted if all significant items (i.e., items with factor coefficients >0.3; Lambert & Durand, 1975) were statistically related with each other. In contrast, within a section, two (or more) subscales (i.e., measures) were extracted if two (or more) sets of significant items were statistically related with each other. Therefore, each emergent scale/subscale yielded a dimension of teaching quality that was represented by the items contained within it. This series of PCAs led to the formation of the following 10 measures (i.e., scales/subscales) of teaching quality perceptions: • Inculcating cultural values • Attitudes towards cultural values • Perceived teacher knowledge and pedagogy • Attitudes towards student-centred learning • Attitudes towards creating a positive classroom environment • Satisfaction with resources and material • Satisfaction with support and opportunity • Classroom teaching motivation • Teaching profession motivation • Attitudes towards diversity All 10 measures represented a 5-point, Likert-format scale/subscale (i.e., 1 = strongly disagree, 2 = disagree, 3 = neutral, 4 = agree, 5 = strongly agree). Each of these measures is described in the following paragraphs, with the classical theory alpha score reliability coefficient and associated confidence interval reported in Table 2. For all individual items that are contained within each of these 10 Likert-format measures, a higher value (i.e., higher rating) indicates greater agreement with the underlying statement. Contrastingly, a lower value (i.e., lower rating) indicates less agreement with the underlying statement (Table 2).
Inculcating Cultural Values Scale. This scale, which contains six 5point, Likert-format items, assesses the degree that the teacher respondent believes that it is necessary for her/him to instill (i.e., inculcate) in her/his students specific values associated with Rwandan culture. Being extracted from the FGDs, the six cultural values that were represented in this scale-one cultural value for each item-were as follows: honesty and integrity, participating in community activities, treating each other fairly, respecting others, forgiving others, and showing tolerance towards others. An example of an item contained in this scale is "I encourage my students to show tolerance towards others." Attitudes Towards Cultural Values Scale. This scale, which also contains six 5-point, Likert-format items, assesses the degree that the teacher respondent believes it is essential that she/he exhibits classroom behaviors that indicates appropriate respect for values that are associated with Rwandan culture. An example of an item that is contained in this scale is "As a teacher, it is important that my students respect me." Perceived Teacher Knowledge and Pedagogy Scale. This scale, containing four 5-point, Likert-type items, assesses the teacher respondent's perception regarding her/his level of knowledge of the subject taught, as well as the teacher respondent's perception regarding her/his pedagogical competence. An example of an item that characterizes teacher knowledge is "I have enough subject knowledge to teach  a It should be noted that whereas for the Phase 1 administration, the teachers were asked to report their current perceptions of teaching quality (e.g., "Indicate the extent to which you agree or disagree with the following statements about being a teacher") before the onset of the COVID-19 context, for the Phase 2 administration, which took place during the COVID-19 pandemic, the teachers were asked to reveal the extent that the COVID-19 context-which included school closures-impacted their perceptions of teaching quality (e.g., "To what extent did school closures affect your views towards the following statements"). my classes well." Also, an example of an item that represents pedagogical competence is "I provide opportunities for my students to identify where they are not understanding." Attitudes Towards Student-Centred Learning Scale. This scale, containing five 5-point, Likert-format items, assesses the degree of the teacher respondent's belief that it is vital for her/him to promote, to nurture, and to facilitate student-centred learning in her/his classroom. An example of an item contained in this scale is "I encourage my students to ask questions." Attitudes Towards Creating a Positive Classroom Environment Scale. This scale, containing five 5-point, Likert-format items, assesses the teacher respondent's perception of her/his level of competence in securing a positive environment for her/his students in the classroom. A representative item contained in this scale is "I encourage students to believe they can do well in their school work." Satisfaction with Resources and Material. This subscale, containing five 5-point, Likert-format items, assesses the degree of the teacher respondent's belief that the resources and materials that are available to her/him are adequate to produce quality teaching in her/his classroom. Two examples of items that are included within this subscale are as follows: "I am satisfied with the availability of textbooks in school for myself and all the children in my class" and "I am satisfied with the amount of material/resources I have access to." Satisfaction with Support and Opportunity. This subscale, containing six 5-point, Likert-format items, assesses the degree of the teacher respondent's belief that she/he has adequate support and opportunities to produce quality teaching. Two examples of items that are included within this subscale are as follows: "I am satisfied with the support from the head teacher" and "I am satisfied with the opportunities for in-service training available to me." Classroom Teaching Motivation Subscale. This subscale, containing five 5-point, Likert-format items, assesses the teacher respondent's level of motivation regarding her/his classroom teaching. An example of an item contained in this subscale is "I have too much material to cover in class." Teaching Profession Motivation Subscale. This subscale, containing seven 5-point, Likert-format items, assesses the teacher respondent's level of motivation regarding the teaching profession as a whole. Two examples of items contained in this subscale are as follows: "I think that the teaching profession is valued in society" and "My work inspires me." Attitudes Towards Diversity Scale. This scale, containing four 5point, Likert-format items, assesses the degree that the teacher respondent has a positive attitude towards her/his classroom reflecting student diversity. Three examples of items included in this scale are as follows: "Students from very poor households have more difficulties than other students," "Students with physical disabilities struggle to grasp learning concepts more than other students," and "Girls have more difficulties than other students in STEM subjects."

Analysis
Descriptive statistics were used to analyse the responses to each scale and subscale (i.e., descriptive research design). In particular, means and standard deviations were computed. Also, a dependent samples t-test was used to compare each of the scales/subscales with each other (i.e., correlational research design).
A second-order PCA was undertaken to explore the structure of each of the 10 measures (i.e., scales/subscales) that was identified via the first-order PCA (i.e., exploratory research design). This analysis produced factors that were extracted via the intercorrelations among the measures, in contrast to a first-order PCA, wherein factors are extracted via the intercorrelations among the original Likert-format items. As such, the second-order PCA represented a higher level of extraction than did the first-order PCA.

Phase 2: COVID-19 and school closures 2.2.1. Research design and participants
The second phase, which took place in March 2021, also involved the use of both a descriptive research design and correlational research design to examine the differences in perceptions of teaching quality, as reported by secondary school STEM teachers in Rwanda, between preand almost exactly 1-year after the onset of COVID-19, as well as to examine the extent to which socio-demographic and locational variables (e.g., gender of teacher, age of teacher, teacher's years of experience, teacher's qualification, teacher's disability status, type of school) varied as a function of perceptions of teaching quality. As stated previously, 1,820 STEM teachers participated in Phase 1 (i.e., February/March 2020). In Phase 2 (i.e., March 2021), 1,931 STEM teachers completed the survey. However, in order to reduce selection bias-which refers to substantive differences between two or more of groups of participants prior to the implementation of the intervention (Onwuegbuzie, 2003), in this case, COVID-19-only the 1,609 STEM teachers who participated (i. e., completed surveys) in both years were included in Phase 2.

Instruments and procedure
Due to the school closure period induced by the COVID-19 pandemic, the survey that was administered face-to-face in Phase 1 had to be delivered via phone in Phase 2. This meant that the Phase 1 survey had to be shortened to avoid over-burdening the participants. As part of shortening the Phase 1 survey, because there were two instruments that measured cultural values (i.e., the Inculcating Cultural Values Scale and the Attitudes Towards Cultural Values Scale), it was decided to eliminate one of these measures-namely, the Attitudes Towards Cultural Values Scale, because the former (i.e., the Inculcating Cultural Values Scale) was more directly applicable to teachers' interactions with students and, therefore, to teaching quality. This removal meant that 9 measures were administered in Phase 2 instead of the 10 measures that were administered in Phase 1. Further, whereas with regard to the Phase 1 survey, which was administered before the onset of COVID-19, the teacher participants were asked to indicate their current perceptions of teaching quality (e.g., "Indicate the extent to which you agree or disagree with the following statements about being a teacher"), for the Phase 2 survey, which was administered during the COVID-19 pandemic, the teachers were asked to indicate the extent that the school closures affected their perceptions of teaching quality (e.g., "To what extent did school closures affect your views towards the following statements").

Analysis
Each item pertaining to all nine measures was recentred. Originally, these items were measured via a 5-point, Likert-format scale/subscale, with responses ranging from 1 (i.e., strongly disagree) to 5 (i.e., strongly agree). Notwithstanding, for ease of interpretation, each item mean within each measure was recentred by subtracting 3 from each response, such that the responses changed from the range of 1-5 to the range of − 2 to 2. Therefore, for each measure, a mean difference below 0 indicated that the COVID-19 context negatively impacted the teachers' perceptions of teaching quality. For these re-centred responses, descriptive statistics-specifically, means and standard deviations-were used to summarise scores pertaining to each scale and subscale for the Phase 2 (i.e., March 2021) responses. Further, the multivariate analysis known as a canonical correlation analysis was employed to determine the relationship between the extent that the COVID-19 context affected the STEM teachers' perceptions of teaching quality and several sociodemographic and locational factors. Broadly speaking, the goal of a canonical correlation analysis is to investigate the multivariate relationship between two sets of variables of interest-a set of independent (i.e., predictor) variables and a set of dependent (i.e., criterion) variables-wherein each set comprises a minimum of two variables (Onwuegbuzie & Daniel, 2003;Thompson, 1984). The aforementioned nine measures of teaching quality perceptions formed the dependent set of variables, whereas the following eight socio-demographic/locational variables served as the independent set: gender (nominal scale; i.e., women vs. men); age (ratio scale); qualifications (nominal scale; i.e., Bachelor's [i.e., undergraduate] degree vs. advanced education diploma); disability status (nominal scale; i.e., reported having a disability vs. reported not having a disability); years of teaching experience (ratio scale); travel time (ratio scale); type of school (nominal scale; i. e., boarding school vs. day school); and size of school (nominal scale; i.e., large and very large vs. small). These variables were selected because of the replicated observation that having a localised knowledge of teaching is crucial both to understanding teaching quality and how to evaluate it (Carter et al., 2022, pp. 1-22;Goe et al., 2008;Mitchell, 2019). This is particularly pertinent within low-income country contexts, given the generalisations and deficit positioning of teachers that often can occur. Moreover, as concluded by Tao (2016), The assumption that teachers-irrespective of gender, years of experience, socio-economic background, religious faith or ethnicity-act and have values based on a singular identity can lead to a deeply misled understanding of teachers. This indiscriminate understanding also leaves subsequent analysis and policies lacking in nuance or relevance. (p. 4).
The eight variables selected comprised six teacher-level variables-namely, gender, age, qualifications, years of teaching experience, travel time from home to school, and disability status-and two systemlevel variables, namely, type of school and size of school. Four of the teacher-level variables (i.e., gender, age, qualifications, years of teaching experience) reflected variables that have been examined in research conducted in both the Global North (e.g., Goldhaber & Anthony, 2007) and the Global South (e.g., Armstrong, 2015). Further, travel time from home to school, the fifth teacher-level variable, was selected because it provided spatial context during the period of school closures. This location variable has been found to be an important predictor of teaching and learning in research conducted examining the effect of COVID-19 at the tertiary level in the Global South (e.g., Onwuegbuzie et al., 2020). Disability status, the final teacher-level variable, was selected because although the protection and provision for students with disabilities represents a key priority of the Government of Rwanda (see, for e.g., Objective 4 of the Education Sector COVID-19 Response Plan), there has been little or no focus on the protection and provision for teachers with disabilities, which was deemed worthy of investigation. The two system-level variables reflected the fact that Rwandan secondary schools are highly stratified (i.e., structurally driven)-with both type of school (e.g., Nizeyimana et al., 2020) and school size (e.g., Bridgeland et al., 2009) being an important focus within Rwandan secondary education.
The number of canonical functions (i.e., variables) that are generated from a canonical correlation analysis is represented by the number of variables in the smaller of the two sets of variables (Thompson, 1984). Consequently, because nine measures of teaching quality perceptions simultaneously were correlated with eight socio-demographic and locational variables, a total of eight canonical functions were produced. Each of these eight canonical functions was tested for statistical significance (i.e., via p values) and practical significance (i.e., via a measure of effect size). Each canonical function is associated with unique standardized coefficients and structure coefficients. For each statistically significant canonical coefficient, the corresponding standardized coefficients and structure coefficients were interpreted in order to facilitate assessment of practical significance (Onwuegbuzie & Daniel, 2003). Table 3 shows that, for the February/March 2020 (i.e., Phase 1) data, the two measures of cultural values (i.e., Attitudes Towards Cultural Values Scale, Inculcating Cultural Values Scale, respectively) generated the most positive ratings (i.e., attitudes). Contrastingly, Table 3 indicates that the Satisfaction with Resources and Material Subscale yielded the least positive ratings (i.e., attitudes). In fact, a series (i.e., n = 45) of paired (i.e., dependent) samples t tests, using a Bonferroni adjustment for Type 1 error (i.e., α = 0.05/45 = 0.0011; cf. Chandler, 1995), revealed that both cultural values scales yielded statistically significantly more positive ratings than did all other measures (i.e., scales/subscales). The Cohen's (1988) d effect sizes associated with these 44 sets of differences (i.e., all 45 differences minus the differences involving the two cultural values measures) ranged from moderate (i.e., d = 0.53; which represented the difference between responses to the Attitudes Towards Creating a Positive Classroom Environment Scale and responses to the Inculcating Cultural Values Scale) to extremely large (i. e., d = 2.51; which represented the difference between responses to the Satisfaction with Resources and Material Subscale and responses to the Attitudes Towards Cultural Values Scale).

Descriptive findings
With regard to the second-order PCA, the Kaiser-Meyer-Olkin (KMO) measure suggested good sampling adequacy, KMO = 0.81 (Field, 2018). Further, the anti-image correlation matrix revealed that all KMO values for the items were greater than the acceptable limit of 0.5 (Field, 2018). Bartlett's test of sphericity indicated that the correlations between the items were sufficiently large for the PCA, Х 2 (45) = 4075.83, p < .0001.

Principal component analysis findings
The second-order PCA suggested the presence of two factors. The pattern/structure matrix stemming from the varimax (i.e., orthogonal) rotation is presented in Table 4. From this table, it can be seen that, using a cutoff correlation of 0.3 recommended by Lambert and Durand (1975), as an acceptable minimum value for pattern/structure coefficients, the pattern/structure matrix (Table 4) revealed that Factor 1 consisted of five items and Factor 2 consisted of four items. Specifically, Factor 1 comprised the following five measures: attitudes towards cultural values, inculcating cultural values, attitudes towards student-centred learning, attitudes towards creating a positive classroom environment, and perceived teacher knowledge and pedagogy (Table 4). In contrast, Factor 2 comprised the following four measures: satisfaction with resources and material, satisfaction with support and opportunity, classroom teaching motivation, and teaching profession motivation (Table 4). Therefore, all nine measures were accounted for by these two factors. Factor 1 explained 32.70% of the variance, whereas Factor 2 accounted for 19.65% of the variance (Table 4). As such, these two factors combined to explain 52.36% of the total variance (Table 4). Interestingly, this proportion of total variance explained was greater than that typically explained in factor analysis solutions (i.e., M = 44.92%, SD = 16.55%, Range = 12.80%-70.20%; Henson et al., 2004). Moreover, this total proportion of variance indicates a large effect size. for all teachers indicates that the majority of secondary school STEM teacher respondents were more inclined to agree or to strongly agree to each of these six items.
In terms of score reliability, Cronbach's alpha pertaining to scores yielded by Factor 1 was 0.83 (95% confidence interval [CI] = 0.82, 0.85) and for scores yielded by Factor 2 was 0.66 (95% CI = 0.63, 0.68). On further examination, Factor 1 indicated the existence of a cultural values and pedagogical component, whereas Factor 2 indicated the existence of a motivational component. More specifically, Factor 1 revealed that teachers who provided high scores (i.e., ratings) with respect to the cultural values and pedagogical component were significantly more likely than were their lower scoring counterparts to indicate the following dimensions of teaching quality: • They expressed a belief in the importance of exhibiting behaviors that demonstrate respect and appreciation for Rwandan cultural values. • They expressed a belief in the importance of instilling in their students cultural values associated with Rwanda. • They expressed a belief in the importance of advancing studentcentred learning via their classroom teaching. • They exhibited very positive perceptions about their ability to create a positive environment that promotes student learning in their classrooms. • They exhibited very positive perceptions about their knowledge of the subject that they teach, coupled with very positive perceptions regarding their levels of pedagogical competence.
Contrastingly, Factor 2 revealed that teachers who provided high scores (i.e., ratings) with regard to the motivational component were significantly more likely than were their lower scoring counterparts to indicate the following teaching quality dimensions: • They reported satisfaction with the resources and materials that they have at their disposal to teach their students. • They reported satisfaction with the support and opportunities that they have at their disposal as teachers. • They reported high levels of teacher motivation with respect to their classroom teaching. • They reported high levels of teacher motivation with regard to the teaching profession within the Rwandan context as a whole. Table 2 presents the score reliability and associated confidence interval for each of the nine measures that were used in both Phase 1 (i.e., February/March 2020) and Phase 2 (i.e., March 2021). Notably, for each of the 9 scales/subscales that were administered during both periods, the reliability coefficient, as measured by Cronbach's alpha, was higher for the Phase 2 data than for the Phase 1 data (Table 2). In fact, as can be seen from Table 2, for the Phase 2 data, the range of the reliability coefficients was from 0.75 (Satisfaction with Resources and Material Subscale) to 0.94 (Inculcating Cultural Values Scale). As such, all of these Phase 2 reliability coefficients in Table 2 exceeded Nunnally and Bernstein's (1994) cutpoint of 0.70. From these findings, it can be concluded that none of the following occurrences adversely affected the psychometric properties of the 9 measures: (a) the changes stemming from the COVID-19 context, (b) the changes in the survey administration between the Phase 1 and Phase 2 administrations (i.e., the survey administered face-to-face vs. the survey administered via phone), or (c) the difference in the stem used for each item in the two surveys (i.e., the teachers being asked to indicate their current perceptions of teaching quality [Phase 1] vs. the teachers being asked to indicate the extent to which the COVID-19 context affected their perceptions of teaching quality [Phase 2]). However, it is not clear why there were discrepancies between the two sets of score reliabilities. Table 5 presents the mean recentred item rating for each of the nine scales/subscales that were administered in Phase 2. Each of these nine measures contains 5-point, Likert-format items such that a higher score on a measure indicates stronger agreement with the items representing each measure. Moreover, for each measure, a mean difference less than 0 indicates that the COVID-19 context negatively affected their sizes within each factor, using a cut-off value of 0.5 recommended by Lambert and Durand (1975). Note: The mean and standard deviation values represent the recentred individual item responses for each measure. Originally, these means reflected a 5point, Likert-type scale, with responses ranging from 1 to 5. However, to facilitate interpretation, each of these means was recentred by subtracting 3 from each response in order to convert the responses such that they ranged from − 2 to 2. Consequently, for each measure, a mean difference less than 0 indicates that the COVID-19 context negatively impacted their perceptions of teaching quality. For instance, for the Inculcating Cultural Values Scale that contains six items, the original mean item rating of 3.70 across all teacher respondents was recentred by subtracting 3.00 such that it generated a recentred mean of 0.70. This 0.70 mean reveals that most teacher respondents tended to report either a positive effect or a strongly positive effect to each of these six items, thereby indicating that the COVID-19 context positively impacted their perceptions of teaching quality pertaining to this scale. Contrastingly, with regard to the Perceived Teacher Knowledge and Pedagogy Scale, the − 0.28 mean indicates that most of the teacher respondents were more apt to report either a negative effect or a strongly negative effect to each item that represent this scale, thereby indicating that the COVID-19 context negatively impacted their perceptions of teaching quality pertaining to this scale.

Effect of the COVID-19 context on perceptions of teaching quality
perceptions of teaching quality, whereas a mean difference greater than 0 indicates that the COVID-19 context positively affected their perceptions of teaching quality. It can be seen from Table 5 that, for four of the nine teaching quality scales/subscales, the responses were negative, with all of them being statistically significantly lower than 0 (all ps < .0001). This finding, in turn, revealed that the COVID-19 context negatively affected their perceptions of teaching quality pertaining to these four elements. These four measures of teaching quality were as follows: Perceived Teacher Knowledge and Pedagogy Scale, Satisfaction with Resources and Material Subscale, Classroom Teaching Motivation Subscale, and Attitudes Towards Diversity Scale-with the Attitudes Towards Diversity Scale generating the most negative responses (Table 5). Moreover, a series of paired samples t tests comparing all the scale/subscale scores with each other, after applying the Bonferroni adjustment, revealed that the Attitudes Towards Diversity Scale responses were statistically significantly more negative than were the other eight scales/subscales, including the next most negative scale/ subscale, Classroom Teaching Motivation Subscale-a difference reflecting a moderately large Cohen's (1988) d effect size of 0.43. This implies that as a result of the COVID-19 context, the teachers reported attitudes towards diversity that were 0.43 standard deviations less negative than were their levels of teacher motivation associated with classroom teaching. In contrast, Table 5 also indicates that the remaining five of the nine teaching quality scales/subscales yielded positive responses, with all responses being statistically significantly higher than 0 (all ps < .0001). This finding, in turn, indicated that the COVID-19 context positively affected the STEM teachers' perceptions of teaching quality associated with the following five elements: the Inculcating Cultural Values Scale, the Attitudes Towards Student-Centred Learning Scale, the Attitudes Towards Creating a Positive Classroom Environment Scale, the Satisfaction with Support and Opportunity Subscale, and the Teaching Profession Motivation Subscale.

Canonical correlation analysis findings
The canonical correlation analysis-which was used to examine the multivariate relationship between the extent to which the COVID-19 context affected the STEM teachers' perceptions of teaching quality and selected socio-demographic and locational variables-revealed that, as a set, the eight canonical correlations were statistically significant at the 5% level of statistical significance (p < .0001). In addition, after removing the first of the eight canonical roots, the other seven canonical roots also were statistically significant (p < .0001; Canonical R c1 = 0.30). These two findings suggested that the initial canonical root was statistically significant and, therefore, provides a meaningful interpretation. Similarly, when the first and second canonical roots simultaneously were removed, the other six canonical roots remained statistically significant (p < .0001; Canonical R c2 = 0.17). However, when the first three canonical functions simultaneously were removed, the other five canonical roots were statistically non-significant at the 5% level. Further, when the first four canonical functions simultaneously were removed, the other four canonical roots were statistically nonsignificant. This lack of statistical significance was repeated for the removal of subsequent canonical roots. As a whole, these findings indicated that, the first two canonical functions were both statistically significant and practically significant (Cohen, 1988); however, the remaining six functions were not statistically significant, thereby rendering any interpretation of them problematic. Therefore, the only two canonical functions that were interpreted were the first two canonical functions. Table 6 presents the findings pertaining to the first canonical root. As can be seen, this table presents both the standardized coefficients (i.e., canonical function coefficients) and the structure coefficients for this first canonical function. Using Lambert and Durand's (1975) cutoff coefficient of 0.3, the standardized coefficients revealed that the variable, type of school, was the only variable that made a notable contribution to the prediction of the dependent variables. With regard to the dependent variables set, satisfaction with resources and material made a noteworthy contribution.
Additionally, the structure coefficients pertaining to the first canonical root indicated that two variables, namely, type of school and travel time, made vital contributions to the first canonical function. The variable, type of school, was by far the greatest contributor. Squaring the  (Lambert & Durand, 1975). Variables that are italicized but not bolded have either a standardized coefficient or a structure coefficient on one or more canonical functions that is practically significant.
Variables that are bolded have both a standardized coefficient and a structure coefficient on one canonical function that are practically significant. Variables that are bolded and italicized have both a standardized coefficient and a structure coefficient on both canonical functions that are practically significant.
structure coefficients indicates the proportion of variance explained by each variable in the independent set of variables. These squared structure coefficients revealed that type of school and travel time explained 88.4% and 10.2% of the variance, respectively. In terms of the set of teaching quality variables, again, only satisfaction with resources and material made a noteworthy contribution, with the square of the structure coefficient indicating that this variable accounted for 81.0% of the variance. As can be seen, both the standardized coefficients and structure coefficients facilitate the identification of which variables from the independent variable set and dependent variable set contribute significantly to the multivariate relationship between the two sets of variables. Conveniently, even more information of the role of the variables can be gleaned by comparing the standardized coefficients and structure coefficients. Table 6 allows such a comparison. This comparison implicates travel time as representing multicollinearity because although the structured coefficient associated with this variable (i.e., .32 ≥ 0.30) was relatively large, the associated standardized coefficient (i.e., 0.18 < 0.30) was relatively small (Onwuegbuzie & Daniel, 2003). In summary, the first canonical root indicated that the multivariate relationship between the extent to which the COVID-19 context affected the STEM teachers' perceptions of teaching quality and selected socio-demographic and locational variables was characterized mostly by the relationship between the type of school (i.e., independent variable) and satisfaction with resources and material (i.e., dependent variable). In addition, travel time played a role in this multivariate relationship. Specifically, teachers in day schools tended to indicate the most negative perceptions of the impact of the COVID-19 context on satisfaction with resources and material. Also, to a lesser extent, teachers with the longest travel time from home to school tended to indicate the most negative perceptions of the impact of the COVID-19 context on satisfaction with resources and material. Table 6 also presents data related to the second canonical function. The standardized canonical function coefficients associated with this second canonical root revealed that both gender and years and experience made notable contributions to the set of independent variables-with gender being the biggest predictor of the set of teaching quality variables. With respect to the set of dependent variables (i.e., teaching quality variables), inculcating values, attitudes towards student-centred learning, perceived teacher knowledge and pedagogy, and level of teacher motivation associated with the teaching profession as a whole (i.e., teaching profession motivation) made important contributions, with level of teacher motivation associated with the teaching profession as a whole making the most noteworthy contribution.
The structure coefficients pertaining to the second canonical function indicated that the following independent variables, namely, years of experience (43.6% explained variance), age (28.1% explained variance), qualifications (9.6% explained variance), and gender (3.6% explained variance), respectively, all made notable contributions. In terms of the set of teaching quality variables, perceived teacher knowledge and pedagogy (32.5% explained variance), level of teacher motivation associated with the teaching profession as a whole (22.1% explained variance), attitudes towards student-centred learning (16.0% explained variance), and satisfaction with resources and material (13.7% explained variance), respectively, made important contributions. Comparing the standardized coefficients and structure coefficients suggested that the following two independent variables were collinear: age and qualifications. Both of these variables were associated with a relatively large (i.e., ≥0.30) structured coefficient (− 0.53, 0.31, respectively) coupled with a relatively small (i.e., <0.30) standardized coefficient (− 0.14, 0.25, respectively). In addition, among the set of teaching quality variables, inculcating cultural values served as a suppressor variable due to its relatively large standardized coefficient (i.e., − 0.31) and relatively small structure coefficient (i.e., − 0.02). Additionally, satisfaction with resources and material indicated multicollinearity (structure coefficient = − 0.37; standardized coefficient = − 0.28). In summary, the second canonical root revealed that the multivariate relationship between the extent to which the COVID-19 context affected the STEM teachers' perceptions of teaching quality and the selected socio-demographic and locational variables was characterized mostly by the relationship between gender and years of experience on the one hand, and attitudes towards student-centred learning, perceived teacher knowledge and pedagogy, and level of teacher motivation associated with the teaching profession as a whole on the other hand. Additionally, inculcating cultural values and satisfaction with resources and material played a role in the multivariate relationship. Specifically, the second canonical root revealed the following: • Male teachers, the youngest teachers, teachers with the least years of experience, and teachers who reported having either Bachelor's degree qualification or no qualification in teaching (as opposed to an Advanced/Postgraduate Diploma in Education/Certificate in Teacher Training), tended to indicate the most negative perceptions of the impact of the COVID-19 context on inculcating cultural values, satisfaction with resources and material, and levels of teacher motivation associated with the teaching profession as a whole. • Male teachers, the youngest teachers, teachers with the least years of experience, and teachers who reported having either Bachelor's degree qualification or no qualification in teaching (as opposed to an Advanced/Postgraduate Diploma in Education/Certificate in Teacher Training), at the same time, tended to indicate the most positive perceptions of the impact of the COVID-19 context on attitudes towards student-centred learning and perceived teacher knowledge and pedagogy.

Discussion
The present investigation is unique inasmuch as it appears to be one of only two dozen studies conducted in the area of teaching quality in Sub-Saharan Africa since 1960, and only the fourth study conducted in this area involving schools in Rwanda, following Iwakuni (2017), Carter et al. (2021), and Dorimana et al. (2022). Moreover, it is the only study to examine the stability of perceptions of teaching quality in the times of a worldwide crisis-namely, COVID-19 and school closures that led to ERTL. As can be seen, this study generated an array of important findings.
A major finding stemming from Phase 1 of this study is that teaching quality can be measured uniquely from the perspective of secondary school teachers in Rwanda. Indeed, this finding supports the contention of Goe et al. (2008) that self-report measures can provide valuable insights that other instruments cannot provide because teachers themselves are the only ones who possess the full knowledge of their capacities and needs. Moreover, Phase 1 validated the themes and subthemes emerging from Carter et al. (2021) qualitative study investigating perceptions of teaching quality among secondary school teachers in Rwanda, and, in turn, supported their conclusions that perceptions of teaching quality in the Rwandan context are multidimensional and multifaceted in nature. This notion of the multidimensionality of teaching quality supports the findings from previous research conducted in the Global North (e.g., Witcher et al., 2001;Gershenson, 2016;Okpala et al., 2009) and the Global South (e.g., Sönmez Boran et al., 2019) on both preservice (i.e., prospective) teachers and inservice teachers.
In fact, the Phase 1 Teacher Survey that emerged from Carter et al. (2021) findings yielded evidence of content-related validity (i.e., the extent to which the items on each of scales/subscales represent the content being measured)-specifically, face validity (i.e., the extent to which the items in each of the scales/subscales appear relevant and important to the teacher respondents), item validity (i.e., the extent to which the items in each of the scales/subscales represent measurement in the intended content area), and sampling validity (i.e., the extent to which the full set of items on each of scales/subscales sample the total content area). The Phase 1 Teacher Survey findings also yielded evidence of construct-related validity (i.e., the extent to which an instrument can be interpreted as a meaningful measure of some characteristic or quality) in general and structural validity (i.e., how well the scoring structure of the instrument corresponds to the construct domain) in particular (cf. Onwuegbuzie et al., 2009). As such, the present findings have made an important contribution to the literature. Interestingly, prior to the COVID-19 pandemic (i.e., Phase 1), both cultural values dimensions (i.e., inculcating cultural values and attitudes towards cultural values) generated statistically significantly more positive responses than did all other measures. This finding is extremely noteworthy because these dimensions are specific to the context of Rwanda, a country that experienced a 4-year civil war that ended in 1994, which led to genocide throughout the country-the aftermath of which led to the implementation of various reforms, with the goal of correcting mistakes made in the past, including curriculum reforms aimed at both primary schools and secondary schools for the purpose of correcting pre-genocide educational imbalances and errors.
Another unique finding that emerged prior to the COVID-19 pandemic is that 9 of these 10 dimensions were found to be subsumed by two components: (a) a cultural values and pedagogical component and (b) a motivational component. As the name might suggest, the cultural values and pedagogical component characterizes an instructional (pedagogical) process that is driven by a cultural-based framework that shapes pedagogy beyond the curricular, instructional, and assessment goals and objectives. This component suggests that, with respect to the context of Rwanda, quality teaching involves incorporating a hidden curriculum-a phrase coined by Jackson (1968), referring to the unwritten, unspoken, unofficial, or implicit cultural messages that are communicated to students while they are in school. In the context of Rwandan schools, this hidden curriculum involves the teacher adopting a pedagogical stance that is culturally relevant (cf. Ladson-Billings, 1992, 2014, culturally responsive (cf. Nelson et al., 2008;Bergeron, 2008;Gay, 2018;Mitchell, 2010;Savage et al., 2011;Sleeter, 2011;Tanase, 2022), culturally competent (cf. Jackson et al., 2018), culturally sensitive (cf. Chen, 2011), culturally proficient (cf. Robins et al., 2006), and/or the like. However, the cultural-based pedagogical framework that is appropriate here is one that is appropriate for the Rwandan context-as measured by the two cultural values measures (i. e., inculcating cultural values and attitudes towards cultural values).
Contrastingly, in the context of secondary school teaching in Rwanda, the motivational component refers to motivation elements-specifically, job satisfaction and teacher motivation-that can be hypothesized as serving either to moderate or to mediate teaching quality. This hypothesis predicts that low job satisfaction and low teacher motivation are predictive of lower perceived teaching quality. At the same time, high job satisfaction and high teacher motivation provide avenues for quality teaching-for example, by developing, promoting, and/or maintaining appropriate cultural values and pedagogical disposition.
As concluded by Watson et al. (2010), Stronge's (2007) conceptualization is "perhaps the most unified and comprehensive" (p. 12). Stronge (2007) identified seven teaching quality components that represent nonacademic and social and emotional behavioral attributes (i.e., teacher affect) that he referred to as the "The Teacher as a Person" category (p. 22). Interestingly, two of these teaching quality components-namely, Caring and Shows Fairness and Respect-form part of the cultural values element aspect of the cultural values and pedagogical component that emerged in the present investigation, while the "Planning and Organizing for Instruction" category-comprising Importance of Instruction, Time Allocation, Teachers ' Expectations,and Instruction Plans (p. 112), combined with the "Implementing Instruction" category-comprising Instructional Strategies, Content and Expectations,Complexity,Questioning,and Student Engagement (p. 113)-map onto the pedagogical aspect of the cultural values and pedagogical component. Further, with regard to the motivation component that emerged in the current study, motivation also is one of Stronge's (2007) seven teaching quality indices that represent the "The Teacher as a Person" category (pp. 110-111). Moving forward, mixed methods research approaches can be used to explore and to test these two sets of teaching quality hypotheses (i.e., a cultural values and pedagogical component and a motivational component) in the context of Rwandan secondary schools. The Phase 2 findings were able to address the research question regarding the effect of the COVID-19 global pandemic and the subsequent closing and reopening of secondary schools (i.e., the COVID-19 context) on the stability of perceptions of teaching quality. Specifically, these findings suggest clearly that, as a result of the COVID-19 context, the secondary school STEM teachers who participated in this study, in general, reported changes in their perceptions of teaching quality. Of the nine teaching quality indicators, these teachers reported that, with respect to four of them (i.e., attitudes towards diversity, levels of teacher motivation associated with classroom teaching, perceived teacher knowledge and pedagogy, and satisfaction with resources and material, respectively), the COVID-19 context negatively affected their perceptions of teaching quality. With regard to the remaining five teaching quality indicators (i.e., inculcating cultural values, level of teacher motivation associated with the teaching profession as a whole, satisfaction with support and opportunity, attitudes towards studentcentred learning, and attitudes towards creating a positive classroom environment, respectively), they reported that the COVID-19 context positively affected their perceptions of teaching quality. Simply put, these findings suggest that the COVID-19 context affected the stability of perceptions of teaching quality. Moreover, these findings provide compelling evidence of the differential impact induced by the COVID-19 context-representing an unexpected situational factor-on perceptions of teaching quality. This differential impact emphasizes further the multidimensional (Carter et al., 2021), multifaceted, and complex nature of teaching quality. In addition, the fact that the STEM teachers' most negative response was with respect to their attitudes towards diversity indicates that the COVID-19 context most negatively affected this aspect of teaching quality. This finding, alongside the finding that levels of teacher motivation associated with classroom teaching was the next most negatively affected, followed, respectively, by perceived teacher knowledge and pedagogy, and satisfaction with resources and material, might reflect the wearing of masks, social distancing laws, and/or other changes which resulted from the COVID-19 rules and regulations that were in place in Rwanda at the time this study took place. Indeed, the period of school closures, wherein teaching had to be conducted remotely (i.e., virtually)-a mode of teaching for which most, if not all, the STEM teachers had little or no experience (Carter et al., 2020)likely made it more challenging for them to teach diverse classes, compounded by a lack of perceived teacher knowledge and pedagogy and a greater need for resources and material, which, in combination, negatively affected their levels of motivation associated with classroom teaching. In particular, the finding that the COVID-19 context negatively affected levels of satisfaction with the resources and materials that were available to them has logical appeal because it likely reflects the fact that teaching remotely necessitated different resources and materials compared with teaching face-to-face at school. On the other hand, the findings that the COVID-19 context positively affected their perceptions of teaching quality pertaining to attitudes towards student-centred learning, attitudes towards creating a positive classroom environment, and satisfaction with support and opportunity are particularly significant, taking into consideration the COVID-19 rules and regulations that existed at the time. For example, compared with teaching face-to-face at school prior to the onset of COVID-19, teaching online amidst (potential) technology difficulties might make it more challenging to create a positive "classroom" environment; so too might teaching while observing social distancing measures and wearing masks.
The Phase 2 findings (i.e., canonical correlation findings) also indicated that five teacher-level variables-namely, gender, age, years of teaching experience, travel time from home to school, and qualifications-were related, either positively or negatively, to changes in perceptions of teaching quality as a direct result of the COVID-19 context. Of particular note is that, as a result of these events, male teachers, the youngest teachers, teachers with the least years of experience, and teachers who report having either Bachelor's degree qualification or no qualification in teaching (as opposed to an advanced/ postgraduate diploma in education/certificate in teacher training), tended to indicate the most negative perceptions with respect to satisfaction with resources and material. Bearing in mind the importance of equity, further research is needed regarding the stability of perceptions of teaching quality for these subgroups of teachers.
That teachers who are male, the youngest, have the least years of teaching experience, and who report having either Bachelor's degree qualification or no qualification in teaching also tended to indicate the most negative perceptions of the impact of the COVID-19 context on levels of teacher motivation associated with the teaching profession as a whole suggests a gender context as well as an age context with respect to motivation levels. Moreover, these findings suggest that the COVID-19 context had the greatest negative impact for these four groups of teachers (i.e., male, young, newest, and least qualified).
Most interestingly, findings revealed that type of school (i.e., day school vs. boarding school)-which is a system-level variable-plays a key role with respect to the changes in perceptions of teaching quality that occurred as a result of the remote teaching that took place within the COVID-19 context. Indeed, this finding, in turn, indicates that these changes in teaching quality perceptions are markedly more related to this system-level variable than to variables that represent the teacher level. That teachers in day schools are more likely to report the most negative perceptions pertaining to satisfaction with resources and material might be reflective of the structural differences between boarding schools and day schools, differences in student-teacher ratios, and differences in resources, among other school-based factors. Regardless, type of school context appears to be a central variable in explaining the impact that the COVID-19 context had on teachers' perceptions of teaching quality and, therefore, should be the subject of additional research to determine how this difference in perceptions manifests itself.

Implications for future research
Another finding of this study was that the COVID-19 context affected the stability of perceptions of teaching quality. However, it is not clear how this COVID-19 context affected this stability. Qualitative data (e.g., semi-structured interviews, FGDs, observations) can play an important role here. Further, bearing in mind the hierarchical nature of the findings, modelling techniques, such as hierarchical level (i.e., multilevel) modelling, could be used to understand better these changes in teaching quality. Also worthy of investigation is the relationship between these changes in perceptions of teaching quality and student learning.

Recommendations and conclusion
The purpose of this two-phase quantitative research study was to build on Carter et al. (2021) qualitative findings, by quantitatively examining perceptions of teaching quality, as reported by STEM teachers, both before the onset of COVID-19 (i.e., Phase 1) and during the COVID-19 global pandemic and the subsequent closing and reopening of secondary schools (i.e., Phase 2). The ability to collect data both just prior to and during the period of COVID-19 and school closures make this study extremely unique. Moreover, these findings have implications for Rwandan educational policymakers (e.g., REB, MINEDUC), Rwandan administrators, teacher training administrators, and, above all, the teachers themselves, as they all seek to maximize teaching quality in Rwandan secondary schools.
In particular, policy makers might consider using the current findings regarding the important role that type of school plays in the teaching quality process to ensure the protection and provision for vulnerable populations, including children with disabilities, girls, and children from lower wealth-quintiles-consistent with Objective 4 of the Education Sector COVID-19 Response Plan. The present findings also could be used to create training initiatives for secondary school teachers that are particularly designed to address aspects wherein they indicated negative changes in perceptions, namely, perceived teacher knowledge and pedagogy, satisfaction with resources and material, levels of classroom teaching motivation, and attitudes towards diversity.
At the school level, school leaders might consider designing and implementing mentoring programs wherein teachers who did not report any negative effects to teaching quality perceptions of the COVID-19related events could mentor those teachers who have been identified-via these score-validated measures-as reporting negative effects on perceived teaching quality. Further, these secondary school leaders might find it fruitful to monitor how these nine sets of negative changes in teaching quality perceptions manifest themselves in the classroom, including in relation to curriculum, teaching, and assessment practices. In particular, it might be worthwhile for these school leaders to monitor the extent to which these negative changes in teaching quality perceptions affect student achievement.
The present study has provided compelling evidence of the importance of monitoring perceptions of teaching quality, especially during times of crises-as has been occurring worldwide since the onset of COVID-19 as a global pandemic and the subsequent school closures. As such, it is hoped that the present investigation motivates other researchers to continue research in this area-especially in Rwanda, so that educational policymakers (e.g., REB, MINEDUC) can make evidence-based decisions and set evidence-based policies.

Funding and/or conflicts of interests/competing interests
This work was supported by Mastercard Foundation (Grant number: RG97056).
The authors have no relevant financial or non-financial interests to disclose. That is, the authors have no competing interests to declare that are relevant to the content of this article; all authors certify that they have no affiliations with or involvement in any organization or entity with any financial interest or non-financial interest in the subject matter or materials discussed in this manuscript; and the authors have no financial or proprietary interests in any material discussed in this article.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. government officials who agreed to take part in this study.