Data Extraction and Preparation. For this study, the author used the open-source R statistic software for data preparation and analysis of PISA data. In R (R Core Team, 2020), the SPSS files in .sav format were read into the R program with the assistance of the haven package (Xxxxxxx & Xxxxxx, 2020). The participants of the survey were the 19,507 anonymized 15-year-old school students from 616 schools throughout Kazakhstan. Kazakhstani school and student data were extracted and merged by “CNTSCHID” which included the common prefix number of “398” for Kazakhstan. Following the selection of the two relevant data sets, they were merged to obtain the complete data for Kazakhstan. This data included 19,507 students from 616 schools and 1,314 variables. Since the “ESCS” was thought to be a significant indicator of student success, it was included in the analyses, along with student gender. Furthermore, after reviewing all student level data relating to “social connections”, 6,725 students and 52 schools were excluded due to missing data and xxxxxxxxx responses (i.e., cases of only one student respondent from a school). This resulted in a remaining 12,782 students nested in 564 schools for the final descriptive statistics for this study. Schools of less than 10 students were also omitted to allow for the potential anomalous effects from an extremely limited numbers of students in schools (Lai & Xxxx, 2014). Therefore, the final data for conducting the main multi-level analyses included 442 schools with 12,044 students. As a result of data preparatory procedures, only the variables of interest in terms of “social connections”, as well as some general school level variables, were chosen for analysis. A total of 58 variables were selected. Both semantically negative questions (i.e., those that ask for levels of student agreement with negatively worded statements) were reverse coded. In Table 1, all of the semantically negative questions are denoted by the letter "R". As a consequence, all of the means for each variable and scale will have an equivalent interpretable meaning.
Data Extraction and Preparation. This author made use of the open-source R statistical software for the data preparation and analysis of the PISA data. In R (R Core Team, 2019), the SPSS files in .sav format were read in with the assistance of the haven package (Xxxxxxx & Xxxxxx, 2020). The participants of the survey were the 19,507 anonymized 15-year-old school students from 616 schools in Kazakhstan. Kazakhstani school and student data were extracted and merged by “CNTSCHID” which included the common prefix number of “398” for Kazakhstan. After reading in both student and school datasets, data were merged. This data included 19,507 students from 616 schools inclusive of 1313 variables. Since the “ESCS” was considered to be an important antecedent of student performance, this variable was selected for analyses as well as gender. However, ESCS was only an individual level variable and not associated with each school. Therefore, there was a need to create average SES for each school and this was defined as “AVG.PISA.ESCS”. After analyzing all school climate student-level variables, 904 students and 190 schools were removed due to missing data and xxxxxxxxx answers (i.e., incidents where there was only one student respondent from a school) reducing the sample to 11,528 students in 426 schools. In order to account for possible anomalous results from very small numbers of students in schools, schools with fewer than 10 students were also removed (Lai & Xxxx, 2014). Therefore, the final data for conducting the main multi-level analyses included 399 schools and 11,317 students. For the current analysis, only the variables of interest concerning school climate as well as some common school-level variables were selected. Overall, 70 variables were chosen. All the semantically negative questions were reverse coded. All the negatively worded questions are earmarked with an “R” in Table 2. Then all of the means for each variable will have the same meaning. Table 2
Data Extraction and Preparation. Data extraction and preparation are crucial steps prior to data analysis. Preparing the data helps to ensure the accuracy and consistency of the values. It also involves handling missing or incomplete data, laying the foundation for the reliable results of the conducted analysis. The open-source R statistical software, an “open-source integrated software that is designed for a variety of arithmetic functions, statistical analyses, and graphical displays” (Gaciu, 2021, p. 14), was used to extract and prepare TALIS 2018 data for this study (R Core Team, 2023). The TALIS 2018 principal and teacher data, formatted in SPSS .sav, were imported into R using the heaven package (Xxxxxxx et al., 2023). This dataset comprised information from 6,566 teachers across 331 schools in Kazakhstan. For principals and teachers, the datasets contained 493 and 448 variables, respectively. Merging these data files was achieved using the “IDSCHOOL” variable with the standardized prefix 398, which denotes Kazakhstan in TALIS 2018 data (OECD, 2019b). The resultant merged dataset encompassed 6,566 teachers and 940 variables. Subsequently, only 45 variables of interest, encompassing school identification (4 items), teachers’ background (6), school climate (13), leadership style (8), professional collaboration (8), workplace well-being and stress (4), and turnover intentions (2), were retained for in-depth analysis. The preparation of the data for analysis involved three sequential steps. Initially, participants exhibiting missing data in the retained variables (n = 699) were identified using the tidyr package (Xxxxxxx et al., 2024) and excluded from the dataset, leading to a sample size of 5,867 teachers. Subsequently, an examination of the number of teachers per school was conducted to identify schools with solitary teachers (i.e., singletons) lacking score variations within the institution. No schools with fewer than four teachers were identified, resulting in no further exclusion of participants in this step. In the final step, schools with no variation in the retained variables were identified and subsequently excluded. A total of 28 schools (out of the original 331) exhibited no variation across teachers' scores in at least one of the variables of interest. Consequently, teachers affiliated with these schools were removed from the sample (n = 412), culminating in a final analytical sample of 5,455 teachers from 303 schools and 45 variables.
