Model Fit Statistics From the Multiple-Group Latent Growth Models of Depressive Symptoms

Modelsχ2dfPRMSEA (95% CI)CFITLIχ2 Difference Test P
Unconditional and conditional models
 1. Unconditional with intercept and slope47.67920.0010.034 (0.022–0.047)0.9620.962
 2. Unconditional with intercept, slope, and quadratic29.52212.0030.035 (0.019–0.051)0.9760.960
 3a. Conditional with sexual minority status freed61.01226<.0010.034 (0.023–0.045)0.9570.950
 3b. Conditional with sexual minority status fixed57.92328.0010.030 (0.019–0.041)0.9630.961.829
 4. Sexual minority status with covariates fixed106.54660<.0010.025 (0.017–0.033)0.9540.946
Single-mediator models
 5a. Family satisfaction: free both paths94.95076.0700.014 (0.000–0.023)0.9780.974
 5b. Family satisfaction: fix both paths93.66678.1090.013 (0.000–0.022)0.9820.979.854
 6a. Peer support: free both paths91.93176.1030.013 (0.000–0.022)0.9770.973
 6b. Peer support: fix both paths92.56278.1250.012 (0.000–0.021)0.9790.976.494
 7a. Cyberbullying victimization: free both paths97.82876.0470.015 (0.002–0.024)0.9720.967
 7b. Cyberbullying victimization: fix both paths103.60778.0280.017 (0.006–0.025)0.9670.962.040
 7c. Cyberbullying victimization: fix path “a” only99.17777.0450.016 (0.002–0.024)0.9710.967.194
 7d. Cyberbullying victimization: fix path “b” only101.44177.0330.016 (0.005–0.024)0.9680.963.035
 8a. Unmet medical needs: free both paths87.82376.1670.011 (0.000–0.021)0.9840.981
 8b. Unmet medical needs: fix both paths87.99278.2060.010 (0.000–0.020)0.9860.984.539
Final multiple-mediator model
 9. Full model with significant mediators137.583116.0840.012 (0.000–0.020)0.9750.970
  • Sex differences in the direct and indirect mediation paths were tested by using χ2 difference tests within a multiple-group analysis framework. "Free" refers to the specified parameters beingfreed across groups. "Fix" refers to the specified parameters being fixed across groups. Mediation paths “a” (predictor to mediator) and “b” (mediator to outcome) in models 5–8 are referred to as "both paths." Nonsignificant χ2 difference tests revealed no significant variation by sex. Because the omnibus χ2 difference test in model 7b was significant, models 7c and 7d were estimated to test whether only 1 or both mediation paths differed across sex. The regression path from cyberbullying victimization to depressive symptoms (model 7d) was more pronounced among male than female participants. CFI, comparative fit index; df, degrees of freedom; RMSEA, root mean square error of approximation; TLI, Tucker-Lewis index; —, not applicable.