CFA-Based Monte Carlo Simulation Results for the Stabilizer Variable Test (SVT): 270,000 Applications Across 540 Conditions

Pre-computed results from the CFA-based Monte Carlo simulation reported in the supplementary material (Section S6) of "Stabilizer Variables in Practice: Applying the SVT to Multi-Group Structural Equation Models" (Yilmaz & Çene, 2026, submitted to Structural Equation Modeling: A Multidisciplinary Journal).
The simulation evaluated the Stabilizer Variable Test (SVT) under a latent variable estimation framework with 270,000 total SVT applications across 540 factorial conditions: 3 moderator configurations (M7, M10, M12) × 4 sample sizes (500, 1000, 1500, 2000) × 5 MI severity levels (0, 0.2, 0.3, 0.5, 0.7) × 5 scenarios (Null, Type A, Type B, Type AB, Moderator) plus 240 graded interaction conditions. Each condition was replicated 500 times with R = 1,000 bootstrap resamples per SVT run. Total computation: 373.9 hours on 14 parallel cores.
Key results: Type I error mean = 0.046 (95% within Bradley bounds), power mean = 0.787 (Type B: 0.813), interaction filter FPR ≤ 0.003 at gamma ≥ 0.15.
Files: CFA_full_simulation_results.rds (R binary, complete replication-level results), CFA_full_simulation_summary_final.csv (condition-level summary, 540 rows).
This is the empirical companion to the theory paper: Yilmaz, S. & Çene, E. (2026). Stabilizer Variables for Measurement Invariance-Induced Heterogeneity: Identification Theory and Testing in Multi-Group Models. Mathematics, 14(6), 1064.
Related Materials:
IsSupplementTo — DOI: 10.3390/math14061064 (theory paper)IsSupplementTo — URL: https://github.com/sy142/stabilizer_practice (code repository)Licence: CC BY 4.0