Creative School Climate as a Link Between Bullying Victimization, School Belonging, and Creative Self-Efficacy: Evidence From PISA 2022 Across Four Countries

Analysis code and documentation for the study examining how creative school climate mediates the serial pathway from bullying victimization through school belonging to creative self-efficacy, using PISA 2022 data from four countries (Turkey, Korea, Finland, Mexico).

Authors

• Özge Yiğit (First author) — Doctoral Candidate, Department of Educational Administration, Faculty of Education, Pamukkale University, Denizli, Türkiye. ORCID: 0009-0000-3534-3474
• Kazım Çelik (Second author, corresponding author) — Prof. Dr., Department of Educational Administration, Faculty of Education, Pamukkale University, Denizli, Türkiye. ORCID: 0000-0001-7319-6567

Citation

Yiğit, Ö., & Çelik, K. (2026). Creative school climate as a link between bullying victimization, school belonging, and creative self-efficacy: Evidence from PISA 2022 across four countries [Analysis code and codebook]. GitHub. https://github.com/Ozgetasmanisa/creativity

License

Code and documentation: CC BY 4.0 (see LICENSE). PISA 2022 raw data are not included (OECD property); see Data Access below.

Data Access

This repository does not contain the PISA 2022 raw data. To reproduce the analyses, download the student questionnaire data file from the OECD PISA 2022 database (publicly available):

• Source: https://www.oecd.org/pisa/data/
• File used: CY08MSP_STU_QQQ.sav (student questionnaire)
• Place it in a local data/raw/ folder (this folder is git-ignored).

The four analytic countries are Turkey (TUR), Korea (KOR), Finland (FIN), and Mexico (MEX); the analytic sample is 26,993 students nested in 891 schools.

Software

• R version 4.5.3 (2026-03-11)
• Key packages:
  • BIFIEsurvey (3.8) — design-based two-level regression with BRR weights
  • lavaan — multigroup confirmatory factor analysis (measurement invariance)
  • dplyr, haven, here

Install:

install.packages(c("BIFIEsurvey", "lavaan", "dplyr", "haven", "here"))

Reproduction Steps

Run the scripts in scripts/ in the following order:

Script	Purpose	Output
2_00_dosya_tasima.R	File/path setup	—
2_01_veri_dogrulama_ICC.R	Data validation, ICC	Table 3 (ICC)
2_02_paralel_mediation.R	Parallel mediation models	Tables 5–6
2_03_seri_mediation.R	Serial mediation (4 equations/country)	Tables 7–8, 02_modeller_seri.rds
2_04_robustness.R	Robustness / sensitivity checks	Tables 9–10
2_05_olcme_degismezligi_final.R	Measurement invariance (MGCFA, CREATSCH)	Table 12
2_06_ulke_farklari_etkilesim_v2.R	Formal cross-national difference tests	Table 13
2_07_moderated_mediation_cinsiyet.R	Index of moderated mediation (gender)	Table 14

All models use:

• Final student weight W_FSTUWT and 80 BRR replicate weights (W_FSTURWT1–80)
• Fay factor = 0.5
• Centering within cluster (CWC) for continuous predictors (Enders & Tofighi, 2007)
• Random-intercept two-level structure (students within schools)

A fixed seed (set.seed(2026)) is used for all Monte Carlo confidence intervals (20,000 draws) to ensure reproducibility.

Repository Structure

.
├── README.md
├── LICENSE
├── CODEBOOK.md            # Variable dictionary and analytic decisions
├── scripts/               # All R analysis scripts (2_00 – 2_07)
└── output/
    ├── tables/            # Generated result tables (CSV)
    └── figures/           # Path diagram and forest plot (PNG/TIFF)

Notes on Measurement

• CREATSCH (creative school climate, 4 items: ST335Q01/Q02/Q05/Q06) was tested for measurement invariance at the item level via MGCFA (WLSMV); full scalar invariance held across the four countries.
• CREATEFF (creative self-efficacy, items ST334) was administered under a rotated (matrix-sampling) design (~2.4 of 5 items per student), so item-level MGCFA was not feasible; the OECD-calibrated WLE index was used (OECD, 2024b).