Okullarda Öğrenme İklimi Ölçeği’nin (OÖİÖ) Geliştirilmesi: Geçerlik ve Güvenirlik Çalışması

Okullarda ogrenmeyi destekleyici bir iklim olusturulmasi, ogretmenin mesleki performansinin gelistirmesi yoluyla ogrenci basarisinin artirilmasinda onemli bir belirleyici olarak gorulmektedir. Bu arastirmanin amaci, ogretmenlerin okullarinda ogrenme iklimine yonelik algilarini olcmeye yonelik gecerli ve guvenilir bir olcme araci gelistirmektir. Arastirma kapsaminda alanyazindan ve ogretmenlerle yapilan gorusmelerden faydalanilarak madde havuzu olusturulmus ve kapsam gecerligi icin uzman gorusune basvurulmustur. Arastirmanin verileri 2020 yilinin bahar doneminde resmi 18 ilkokul, 20 ortaokul ve 15 lisede gorev yapan gonullu 589 ogretmenden olusan iki farkli calisma grubundan toplanmistir. Verilerin cozumlenmesinde madde analizi, acimlayici faktor analizi, birinci ve ikinci duzey dogrulayici faktor analizi yapilmis ve guvenirlik katsayilari hesaplanmistir. Madde analizi ve AFA sonucunda toplam 22 madde ve dort alt boyuttan olusan bir olcek elde edilmistir. Ortaya cikan alt boyutlar isbirlikci ortam, okul muduru destegi, okul imkânlari ve mesleki ilgi olarak adlandirilmistir. DFA sonucunda elde edilen bulgular olcegin kabul edilebilir bir uyuma sahip oldugunu gostermistir. Olcegin guvenirligi icin .93 olarak hesaplanan Cronbach Alfa ic tutarlik katsayisi olcegin oldukca guvenilir oldugunu ortaya koymustur. Bu bulgulara gore, Okullarda Ogrenme Iklimi Olcegi’nin ogretmenlerin ogrenme iklimine yonelik algilarini olcmek amaciyla kullanilabilecek gecerli ve guvenilir bir olcme araci oldugu soylenebilir. Gelecek calismalarda olcegin ozel okul ogretmenleri uzerinde gecerlik ve guvenirlik calismalarinin yinelenmesi onerilebilir.


Introduction
In fast-changing, complicated, and diversified working environments, employees need to develop their competencies to keep up with current conditions. According to Nikolova et al. (2014), organizations' survival depends on creating encouraging environments for learning.
Providing a learning environment in which employees can improve their competencies and share with their colleagues is an effective way of accomplishing organizational goals (Marsick & Watkins, 2003). Employees' beliefs and attitudes towards learning are based on work context features such as creativity, innovation, supporting professional development, and administrative practices (Bates & Khasawneh, 2005). Besides motivation and learning skills, employees' perceptions of the work environment also affect their learning (Atwal, 2013). It is asserted that these individual perceptions of the workplace are examined under the name of "climate" and are determinant upon creativity, learning, and performance (Hetland et al., 2011;Rousseau, 1988). These views saliently yield to highlight the need to create a positive learning climate to accomplish learning effectively.
In recent years, learning climate has received wide scholarly attention as an essential factor in enhancing teacher effectiveness (Eldor & Harpaz, 2016). Furthermore, creating a climate that supports learning is seen as an effective way of promoting the employees' professional growth (Nikolova et al., 2014). Furthermore, the existing literature revealed the empirical link between learning climate and teacher professional learning (Atwal, 2013, Opfer, Pedder & Lavicza, 2011, learning capacity (Prieto & Revilla, 2006), willingness to use digital learning materials (Vermeulen et al., 2017), job satisfaction and motivation (Shoshani & Eldor, 2016), innovative behaviors (Sung & Choi, 2014), student learning (Bowen & Kilmann, 1975;Moreland, 1984). A growing body of research has debated that organizational conditions can be a supportive or hindering factor for learning (Cortini, 2016;Nikolova et al., 2014). Learning climate as a vital component of teacher learning is seen as an environment consisting of various opportunities and activities through which teachers acquire, share, and learn knowledge (Shoshani & Eldor, 2016). Learning climate as an organizational factor plays a crucial role in reaching formal and informal professional learning opportunities (Atwal, 2013). In schools with a positive learning climate, all school members solve educational problems and take responsibility for a common goal. Teachers who perceive a positive learning climate feel more valued and show more willingness to participate in learning activities (Şişman, 2011). A positive learning climate contributes to teachers' using alternative teaching practices and reflecting on experiences by arousing feelings of success and satisfaction (Shoshani & Eldor, 2016). Furthermore, a positive learning climate is seen as a pivotal factor in enhancing teachers' professional competencies (Deal & Peterson, 1999).
Since improving, teacher quality by supporting teacher learning is vital for schools to enhance both teacher and student performance (Hodkinson & Hodkinson, 2005;Shoshani & Eldor, 2016), developing teacher capacity and increasing their professional competencies has become the focal point of educational policies (Villegas-Reimers, 2003). Norms, patterns, and practices of schools can be a supporting or hindering factor for teacher learning. Schools that provide continuous learning opportunities, fostering research, dialogue, cooperation, and teamwork, and create a system in which leadership is distributed highly contribute to teacher learning (Opfer, Pedder & Lavicza, 2011). In teacher learning, which is an interactive and continuous exchange of knowledge and experience, school and classroom are the settings where learning occurs (Liu, Hallinger & Feng, 2016). Effective schools support teachers and provide a positive atmosphere for learning (Schunk & Pajares, 2009). However, teachers who do not have supportive learning conditions in their schools can have difficulties in learning new things and transferring them to their teaching practices (Hollingsworth, 1999).
Learning climate is claimed to have gained little scholarly attention due to the lack of valid and appropriate scales (Nikolova et al., 2014). As for Turkey, it is seen in the existing literature that scales adapted from different languages and cultures are used for measuring the learning climate. In some studies, the "Learning Climate Survey" developed by Tennessee School Safety Center and Lincoln Schools were used (Bora, 2010;Çamur, 2006;Şentürk & Mutlu, 2019). In another study, Bozdoğan (2010) used "The Learning Climate Inventory" developed by Hoyle (1972) by adapting it into Turkish. Considering every country has a different educational system and cultural construct, a valid and reliable scale for measuring teachers' perceptions of learning climate in schools appropriately developed for Turkish culture is required.
In the existing literature, it is seen that studies on learning climate in schools mostly aimed to measure perceptions of the students (Bama, 1999;Damico & Roth, 1993;Gerold & Barnes, 1986;Kanadlı & Bağçeci, 2016;Strodl, 1988). According to Social Cognitive Theory, separate roles of students and teachers have led to differentiation in their school context perceptions (Bandura, 1991). The school climate components, which seem less meaningless for students, can have a substantial effect on teachers' perceptions. These perceptions affect teachers' decisions on teaching (Perry & Rahim, 2011). In this regard, it seems essential to develop new scales to measure teachers' perceptions of school climate (Liu et al., 2014).
Meanwhile, though the related literature includes tools measuring learning climate out of educational organizations (Eldor & Harpaz, 2016;Bates & Khasawneh, 2005;Nikolova et al., 2014), there is a limited number of scales measuring school climate-related to teacher learning (Shoshani & Eldor, 2016). Besides, although school learning community scales are also used for investigating the school context of teacher learning, some dimensions of these scales are not appropriate for the Turkish educational system. It is, therefore, suggested that new measurement instruments should be developed by considering the cultural structure and codes of the Turkish educational system (Öğdem, 2015).
Developing teacher learning requires defining features of the learning climate in schools. It is hypothesized that developing a valid and reliable scale for measuring teacher learning climate will be beneficial both theoretically and practically. Furthermore, this study's learning climate scale is thought to contribute to the existing literature by enabling investigation and improvement of learning climate in schools.

Learning Climate
Learning climate is a relatively more recent concept emerged as a result of research on effective schools and focusing specifically on learning that separates it from the terms of school climate and organizational climate (Resendiz, 1994). During the period of restructuring of schools, the learning organization approach was a pivotal factor for the emergence of the learning climate (Çamur, 2016). Furthermore, the relationship of learning climate with learning quality and organizational performance made the term more widely accepted (Mikkelsen & Grønhaug, 1999). According to Marsick and Watkins (2003), learning climate refers to employees' perceptions of beneficial organizational activities that enable them to generate, learn, and transfer knowledge. Lowe (1990) defined learning climate as feelings and perceptions of individuals towards physical and psychological variables that enable a positive learning atmosphere. According to Honey and Mumford (1996), the learning climate is an environment where behaviors and practices towards professional development are supported. Nikolova et al. (2014) described learning climate as employees' perceptions of organizational policies and practices that develop, reward, and enhance their learning behaviors. Similary, Eldor, and Harpaz (2016) stated that learning climate is a concept that reflects perceptions of employees of the degree to which the atmosphere of the organization encourages learning. Lezotte et al. (1980) stated that the learning climate in schools has a distinctive meaning. By combining the terms "school" and "learning climate," an atmosphere comprised of ongoing learning practices in schools is underlined. A positive school climate that facilitates learning and teaching is a crucial component of learning culture in schools. In schools with a positive learning climate, teachers collaborate with their colleagues and school administrators, respect, and trust each other (Ellis, 1988). Silins, Mulford, and Zarins (2002) illustrated that in schools with a positive learning environment, leadership is distributed among school members, collaboration is encouraged, and teachers' learning needs are fulfilled. Kaplan and Geoffroy (1990) claimed that a positive learning climate helps teachers take more responsibility and risks for teaching by strengthening their sense of professional competency. Teachers in a supportive learning climate feel more effective, commit themselves to school goals, and trust their colleagues (Liu et al., 2014). A positive learning climate promotes communication among teachers, fosters their motivation, and makes them more effective in teaching (Yielding, 1993). When teachers perceive a supportive learning climate, they tend to take more responsibility for student learning, show a willingness to teach, and never give up against problems (Sweetland & Hoy, 2000).

Dimensions and Measurement of Learning Climate in Schools
One of the pioneering research on learning climate in schools was conducted by Hoyle (1972), who developed "The Learning Climate Inventory" to determine teachers' perceptions of learning climate. The scale consisted of five sub-dimensions as leadership, freedom, evaluation, compliance, and cooperation. Moreland (1984) developed "The School Learning Climate Assessment Instrument" for investigating teachers' perceptions of the learning climate. In this research, schools' learning climate was dimensioned as instructional leadership, academic emphasis, safe environment, expectations/evaluation, and test usage.
Similarly, "Learning Climate Questionarrie" developed by Bartram et al. (1993) involved seven dimensions as management style, time, autonomy and responsibility, team style, opportunities to develop, guidelines, and satisfaction. Another scale used to measure schools' learning climate is the "Learning Organization Questionnaire" developed by Marsick and Watkins (2003). It was used in several studies to investigate schools' learning climate (Eldor & Harpaz, 2016;Shoshani & Eldor, 2016). The scale has a single structure construct involving items related to continuous learning, inquiry, dialogue, team learning, collective vision, and strategic leadership. Aldridge, Laugksch, and Fraser (2006) developed the "School Level Environment Questionnaire" to assess teachers' perceptions of the schoollevel environment. The measurement instrument involved seven dimensions: parental involvement, student support, collegiality, familiarity, innovation, resource adequacy, and work pressure. Nikolova et al. (2014) developed the "Learning Climate Scale" to measure employees' perceptions of organizations' learning climate. The scale involved three factors as facilitation climate, appreciation climate, and error avoidance climate. Another scale that is used widely in the literature is the "Learning Climate Survey" developed by Tennessee School Safety Center and Lincoln Schools. The scale includes dimensions related to safety, facilities, sources and environment, participation in decision making, student management, curriculum, teaching, success, and morale.
In the existing literature, it is seen that the learning climate in schools includes various sub-dimensions. However, definitions and findings from several studies explicitly revealed that the learning climate involves some common points. These common factors are collaboration, sharing, sources, personal characteristics, and a common vision. Highlighted common dimensions served as a guideline for this research in developing a scale for measuring the learning climate in schools required for effective teacher learning. The dimensions of the scale developed in this study were regarded to be consistent with the dimensions revealed in previous studies in the literature. As many of the earlier scales of learning climate were not developed for measuring perceptions of teachers, teacher views were also used in the process of item preparation.

Method
Research data were gathered under the permission of the Hacettepe Üniversitesi Ethics Commission granted on 09 th June 2020 with document number 35853172-600. This study, which aimed to develop a valid and reliable scale for measuring teachers' perceptions of learning climate, was conducted through a descriptive scanning model. A descriptive model is a research approach that aims to describe a previous or recent existing condition as it is (Karasar, 1999). In the following part, information on study groups and the data collection process is presented.

Participants
Within this study, data were collected from two separate study groups. The study groups consisted of volunteer teachers employed in state elementary, secondary, and high schools in the spring term in 2020. In determining the sample, the convenience sampling method was used. The convenience sampling method is claimed to provide the researcher pace and practicality since the researcher prefers a situation that is closer and more accessible (Yıldırım & Şimşek, 2008). Exploratory Factor Analysis (EFA) was performed with the online data gathered from the first study group consisting of 279 teachers, and CFA was performed with the data gathered from the first study group consisting of 310 teachers. The sample size is claimed to be essential for strong factor analysis (Tabachnik & Fidell, 2007).
In the existing literature, there are various views related to the sample size required for factor analysis. For example, Kline (1994) stated that the required sample size for factor analysis should be at least 100, but greater would yield more good results. According to Comrey and Lee (1992), 200 participants are average, and 300 participants are good sample size for factor analysis. Cattell (1978) asserted that a sample size from three to six times the item number or approximately 250 is sufficient for factor analysis. According to Tavşancıl (2006), the sample size should be at least five times the item number for strong factor analysis. Considering the suggested standards in the literature, it can be stated that the sample size is sufficient for performing factor analysis in this study. Demographic data related to the study groups are given in Table 1.  Table 1 is examined, it can be seen that of the participants in the first study group %59.5 (n=166) were females and %40.5 (n=113) were males. Of the participants in the first study group %, 36.2 (n=101) were primary school teachers, %35.8 (n=100) were secondary school teachers, and %28 (n=78) were high school teachers. Distribution of teachers by years of experience shows that of the participants %13.6 (n=38) have 1 to 5 years, %21.5 (n=60) have 6 to 10 years, %30.8 (n=86) have 11 to 15 years, %17.6 (n=49) have 16 to 20 years, and %16.5 (n=46) have 21 years and over teaching experience. Of the participants, %77.1 (n=215) completed graduate and %22.9 (n=64) completed undergraduate education.

The Process of Scale Development
In the process of determining items, deductive and inductive methods were used in combination. In the deductive approach, items of the scale are determined as a result of a detailed analysis of the existing literature. In the inductive approach, items are determined after interviews with a group defined in the target population (Hinkin, 1995). Within the scope of this study, the existing literature related to the learning climate was scrutinized.
Scales of learning climate developed by Nikolova et al. (2014), Opfer, Pedder, andLavicza (2011), Bartram et al. (1993), andHoyle (1972) were also examined. Furthermore, a total of 12 teachers were interviewed to evaluate their views on the learning climate. As a result of the literature review and interviews with teachers, an item pool consisting of 35 items was created. The scales were thought to be represented under four dimensions: principal support, collaborative environment, school facilities, and professional support. For example, items were stated as "School principals encourage teachers to learn new things" in the principal support dimension; "Teachers share their professional knowledge and experience" in the collaborative environment dimension; "Teachers own necessary sources (books, magazines, and internet) for professional learning" in school facilities dimension and "Teachers follow recent developments in their fields" in professional interest dimension. To determine the degree of frequency, the statements were designed as a 5-point Likert-type scaling ranging from 1=Never to 5= Always.
Items of the scale are required to be sufficient in reflecting the intended behaviors.
Accordingly, one of the methods for evaluating content validity is appealing for expert opinions (Büyüköztürk et al., 2016). Expert opinions should be benefitted for determining whether items under each factor represents behaviors in that factor in multi-dimensional measurement instruments (DeVellis, 2003). Therefore, the items were presented to 13 field experts, 12 of them are employed in Educational Administration, and one is in Measurement and Evaluation Departments. Expert opinions were collected through a form consisting of options as "Essential," "Partially essential," and "Remove." Participants were requested to state their opinions for the items they evaluated as partially essential and not essential.
Analysis of expert opinions was conducted via the Lawshe method. In this method, five to forty experts are required for determining content validity. By adding up the experts' responses related to each item, a content validity ratio (CVR) is obtained. The minimum CVR value is calculated as .54 for 13 experts (Yurdugül, 2005). In this study, therefore, CVR values were calculated for each item. Nine items below .54 were removed from the form. Besides, according to the suggestions of experts, statements of eight items were revised. To evaluate the items in terms of suitability for language, two Turkish language teachers were consulted. As a result of experts' evaluation in terms of spelling rules, expression, and punctuation, items of the scale were re-examined. Furthermore, 10 teachers out of the study groups were asked to evaluate the comprehensibility of the items. Regarding the feedbacks, the scale consisting of 26 items was made ready for application.

Data Analysis
Data were analyzed via SPSS 20 and AMOS 23 statistical programs. In the data analysis process, firstly, Z scores, and Mahalanobis distance values were calculated to arrange data for factor analysis. Item analysis was performed to investigate whether the data is relevant to the desired attitude and can discriminate different degrees (Tezbaşaran, 2008). The normality of the data was checked through the analysis of Skewness and Kurtosis values. To determine the multicollinearity between items of the scale, correlation coefficients (r < .90) were examined (Tabachnick & Fidell, 2007). Item discrimination was evaluated through item-total correlation and upper and lower %27 group scores.
Exploratory Factor Analysis (EFA) was performed to examine the construct validity of the measurement instrument. EFA is an analysis used to determine factors based on the relationship between variables and used widely in studies of scale development (Büyüköztürk, 2012;Field, 2013). Principal Components Analysis (PCA) was used as a factor extraction method since it is claimed to be more effective in reducing factor uncertainty and psychometrically a more powerful method (Tabachnick & Fidell, 2007).
Direct Oblimin as one of the oblique factor rotation methods was preferred due to the correlations between factors (Çokluk, Şekercioğlu & Büyüköztürk, 2010). Before EFA, Kaiser-Meyer Olkin (KMO) value was applied, and Bartlett's Sphericity test was performed to examine whether the data met the criteria for factor analysis. The factor structure was determined by considering an eigenvalue greater than one and scree-plot.
To confirm the four-factor structure that emerged from EFA, first order and second order Confirmatory Factor Analysis (CFA) were conducted. CFA is a factor analysis method used to test a structure or hypothesis of a previously emerged relationship (Field, 2013 Cronbach's alpha and correlation coefficients between factors were calculated for evaluating the reliability of the instrument.

Findings
This section presents findings related to item analysis, exploratory factor analysis, first and second-order confirmatory factor analysis, and reliability analysis.

Item Analysis
When examining Table 2, which displays the item analysis results for examining relevancy of the data with the desired attitude and discriminability of the items, it is seen that arithmetic means of the items varied between 2.78 and 3.96. Considering Skewness and Kurtosis values, it was found that item 12 did not meet the univariate normality assumption.
Skewness and Kurtosis values between ±1 exhibit that the data does not pose a significant deviation from normality (Çokluk, Şekercioğlu & Büyüköztürk, 2010). It is suggested that items deviating from these values can be excluded from the instrument in the scale development process (Şencan, 2005). Item 12 was, therefore, removed from the instrument.
Skewness and Kurtosis values of the remaining items were found to vary between .037 and -.953. These values prove the normal distribution of the research data.
Positive and higher item-total correlation values indicate that an instrument's items illustrate similar properties, and internal consistency is higher. In Likert-type scales, items that exhibit lower relationships with the total score are suggested to be removed from the scale (Tezbaşaran, 2008). Items having above .30 item-total correlation is stated to be sufficiently discriminable (Büyüköztürk, 2012). Therefore, item 10 was removed from the scale as it had a below .30 item-total correlation value. The analysis was re-performed after excluding these items, and the remaining items were calculated to range between .47 and .69. Lastly, t values regarding the differences between upper and lower %27 group scores were found to vary between 8.06 and 15.89 and significantly differed in p<.05 level.
According to these findings, it can be stated that each item in the scale is sufficiently discriminable.  (Büyüköztürk, 2012;Field, 2013).
According to these findings, it was seen that research data were suitable for factor analysis.
Principal Components Analysis as a factor extracting method was used in EFA. The results of the analysis revealed that items of the scale were gathered under four factors, with an eigenvalue greater than one. Findings related to the factor structure of the scale were displayed in Table 3. In determining factor numbers, eigenvalue, variance rate, and scree-plot are seen as important indicators (Büyüköztürk, 2012;Çokluk, Şekercioğlu and Büyüköztürk, 2010).
When the Table is examined, it is seen that four factors were extracted with an eigenvalue of 9.760, 3.342, 1.749, and 1.262, respectively. In factor analysis, factors having an eigenvalue greater than one are accepted as significant factors (Field, 2013). This finding of the study indicated that the scale consisted of four factors with an eigenvalue greater than one.
Factor number that firstly explained 2/3 of the total variance is seen as significant. Higher explained variance means a better measurement of the related structure (Büyüköztürk, 2012). This finding indicated that the total variance of LCS was sufficient. Scree plot graph used one of the important indicators in determining factor structure is given in Figure 1.  (Çokluk, Şekercioğlu, & Büyüköztürk, 2010). When Figure 1 is examined, a decline in the slope is observed as of the fourth factor, and the graph begins to hold a horizontal structure.
It can be, therefore, argued that the scale had a four-factor structure.
To determine the factors under which the items were gathered, Direct Oblimin as an oblique rotation technique was preferred since the factors were assumed to be associated with each other (Field, 2013). As a result of EFA, factor loadings and common factor variances are shown in Table 4. It is argued that item factor loadings should be above .50 to ensure construct validity requirements (Fornell & Larcker, 1981;Nunally, 1978). Therefore, .50 was specified as a lower limit of factor loadings. As a result of the analysis, the factor loadings of a few items were calculated to be lower than .50. For this reason, items 14 and 21 were removed from the scale, and the analysis was re-performed. No items were found to have high factor loadings in more than one factor. Besides factor loadings, common factor variances are also used in evaluating the findings of EFA. Common factor variances show the degree to which the extracted factors are jointly represented. Therefore, items having a common factor variance below .20 must be excluded from the instrument (Şencan, 2005). The analysis revealed that there were no items that have a common factor variance below .20.
Consequently, when Table 4 is examined, it could be seen that factor loadings of the items varied between .620 and .859. Costello and Osborne (2005) stated that items with factor loadings above .50 have strong load values. It is also seen that all items sufficiently contribute to the common variance. Four dimensions obtained as a result of EFA were entitled considering the content of the items. Sub-dimensions and items within these subscales are displayed in Table 5. When Table 5 is examined, six items (15,16,17,18,19,20) under "Collaborative Environment", seven items (1,2,3,4,5,6,7) under "Principal Support", five items (22,23,24,25,26) under "School Facilitites" and four items (8,9,11,13) under "Professional Interest" sub-dimensions.

First Order Confirmatory Factor Analysis
To validate the structure consisting of 22 items and four factors and test model-data fit, firstorder CFA was performed with the data gathered from 310 participants. As a result of the analysis, the t values of the variables were examined. In the literature, it is stated that t values greater than 1.96 and 2.56 are significant at .05 and .01 levels respectively, and items that are not significant should be removed from the scale (Çokluk, Şekercioğlu & Büyüköztürk, 2010). For the existing four-factor model, t values were found to vary between 9.34 and 18.86. This finding indicated that all items were significant at the .01 level and not required to be excluded from the scale. Afterward, fit indexes of LCS were examined. The findings of the first-order CFA are displayed in Table 6.  Table 6 is examined, it is seen that Chi-square and degree of freedom values were calculated to be χ2=434, df=203. When the values were rated, the result of χ2/df=2.14 was obtained. χ2/df value below 3 and 5 indicates perfects fit and medium-level fit, respectively (Jöreskog & Sörbom, 1993). Therefore, it could be stated that the χ2/df value estimated for this model was an indicator of a good model fit. RMSEA value, which is one of the widely used fit indexes in DFA, is claimed to imply a good fit below .05 value and an acceptable fit below .08 value (Browne & Cudeck, 1993;Kline, 1994). RMSEA value calculated as .06 for this model was an indicator of adequate fit. RMR value represents a good fit and acceptable fit below .05 and .10, respectively. RMR value found to be .05 for this model indicated a good model fit (Hu & Bentler, 1995 (Çokluk, Şekercioğlu & Büyüköztürk, 2010;Schumacker & Lomax, 2004;Sümer, 2000). However, since GFI and AGFI values are affected by sample size, values above .85 are claimed to be acceptable in evaluating model fit (Anderson & Gerbing, 1984;Bryant, Yarnold & Grimm, 1996;Schermelleh-Engel & Moosbrugger, 2003). In the light of such information in the existing literature, it can be stated that CFI values indicated a good fit; NFI, TLI, GFI, and AGFI values indicated an acceptable fit. The path diagram of the four-factor model and factor loadings obtained from CFA are shown in Figure 2.
When standardized coefficients are examined in Figure 2, it is seen that the correlation coefficients of variables vary between .42 and .75. Factor loadings of the items were calculated between .72 and .86 for "Collaborative Environment," .80 and .88 for "Principal Support," .61 and .73 for "School Facilities," and .73 and .85 for "Professional Support" sub-dimensions. All items having factor loadings above .50 indicated that each item excellently represented the factor to which it belongs (Fornell & Larcker, 1981

Second-Order Confirmatory Factor Analysis
According to Meydan and Şeşen (2011), multi-factor scales should also be tested in terms of second-order multi-factor model fit. In this regard, second-order CFA was conducted to determine whether previously obtained four dimensions were converged under a more upper and inclusive factor, "Learning Climate." The goodness of fit indexes calculated as a result of second-order CFA performed by adding the second-order latent variable (learning climate) to the previously obtained model with four latent and 22 observed variables is given in Table 7.  (Anderson & Gerbing, 1984;Jöroskog & Sörbom, 1993;Kline, 1994;Schumacker & Lomax, 2004). Path diagram which shows the factor-scale relation is displayed in Figure 3. Furthermore, according to second-order CFA results, t values were calculated to be 9.52 for a collaborative environment, 7.62 for principal support, 7.41 for school facilities, and 9.53 for professional interest sub-scales, which meant the result was significant at p<.01 level.
The significance of the T value is an indicator of the model's acceptability (Schumacker & Lomax, 2004). In the light of these findings, it can be concluded that LCS consisting of 22 items and four-factor is valid and has adequate goodness of fit.

Reliability Analysis
According to Tezbaşaran (2008), internal consistency should be tested in Likert-type scales, and calculating Cronbach's Alpha method is the most appropriate way of testing reliability.
For this reason, to determine the reliability level of the sub-dimensions and the whole LCS, Cronbach's Alpha coefficients were calculated (see Table 9). When Table 9 is examined, it is seen that Cronbach's Alpha coefficients for subdimensions of LCS varied between .80 and .91. Cronbach's Alpha coefficient for the total scale was calculated as .93. Cronbach's Alpha value greater than .70 is claimed to be sufficient in terms of an instrument's reliability (Fraenkel & Wallen, 2006;Nunnally, 1978;Şencan, 2005). It can be, therefore, stated that LCS is a reliable scale. Furthermore, correlations of sub-dimensions with each other and the total scale were evaluated to test the scale's internal consistency (see Table 10). When Table 10 is examined, it is seen that correlation of sub-dimensions with each other, and the total scale was significant at p<0.01 level. Correlation coefficients between the sub-dimensions of the scale varied between .31 and .65. In the literature, very high and low correlations between factors are not the desired situation (Çokluk, Şekercioğlu & Büyüköztürk, 2010). Meanwhile, correlation coefficients below .80 indicate that the scale's discriminant validity is ensured (Brown, 2006). In the light of current findings, it can be concluded that positive and medium-level correlations between the sub-dimensions of the scale show that each sub-dimension measures different features. Furthermore, high and significant relationships were detected between sub-dimensions and total scale. This finding was another indicator of internal consistency and reliability of the scale.

Discussion and Conclusion
This research aimed to develop a valid and reliable scale that measures teachers' perceptions of the learning climate in schools. Accordingly, the existing literature related to learning climate was reviewed, and previous measurement tools of learning were investigated in detail (Bartram et al. 1993;Hoyle 1972;Nikolova et al. 2014, Opfer, Pedder & Lavicza 2011. It was predicted that the scale could consist of four dimensions based on the existing literature. For testing validity, item analysis, exploratory factor analysis, first and secondorder confirmatory factor analysis were used, and reliability was tested through Cronbach's Alpha coefficients. EFA was performed with the online data gathered from the first study group consisting of 279 teachers, and CFA was performed with the data gathered from the first study group consisting of 310 teachers. As a result of the analysis, a scale consisting of 22 items and four factors were obtained. The four dimensions were entitled of a collaborative environment, principal support, school facilities, and professional support. The first-order and second-order DFA results determined whether the four sub-dimensions were combined under the upper learning climate factor indicated that the four-factor model is confirmed and has acceptable-level construct validity (Jöroskog & Sörbom, 1993;Hu & Bentler, 1995;Schumacker & Lomax, 2004). In order to determine the reliability of the LCS, Cronbach's Alpha coefficients were calculated. Cronbach's Alpha coefficients were found as .93 for the scale as a whole, .91 for a collaborative environment, .94 for principal support, .80 for school facilities, and .86 for professional interest sub-dimensions. In this regard, it can be argued that LCS is a reliable measurement tool (Fraenkel & Wallen, 2006).
Lastly, correlations between the sub-dimensions were calculated between .31 and .65.
Correlation coefficients between the factors below .80 is an indicator of discriminant validity (Brown, 2006).
In conclusion, the results of the previously stated analyses revealed that LCS is a valid and reliable scale that can be used to measure teachers' perceptions of the learning climate in schools. The scale consisting of 22 items and four factors is a five-point Likerttype (from never to always) scale. The lowest and higher scores that can be received are 22 and 110, respectively. The increase in the received scores indicates the increase in teachers' perceptions levels of the existing learning climate in their schools. The scale developed within this study is thought to contribute to the research conducted to determine the characteristics of schools. It is suggested that the scale's validity and reliability should be tested in different samples, such as private school teachers.

Limitations
This research has some limitations. Firstly, the validity of the scale was tested through exploratory and confirmatory factor analysis. A criterion-based validity was not tested. This limitation can be removed in further studies. Secondly, data of the study were gathered from teachers employed at public schools. Considering the different learning facilities between state and private schools, this study can be a limitation not to include the characteristics of the learning climate in private schools. Therefore, it can be proper to re-test the validity and reliability of the data gathered from private school teachers in further studies.