Skip to main content
. 2019 Feb 4;3(5):nzz010. doi: 10.1093/cdn/nzz010

TABLE 5.

Comparison fit statistics between ML and QML with Satorra-Bentler correction among model iterations

Iteration 1 Iteration 2 Iteration 3 Iteration 4 Iteration 5
Complete cases without weightings (n = 10,574) Complete cases with weightings (n = 10,574, total sample = 153,038,278) Imputed complete cases with weightings (n = 10,574, total sample = 153,038,278) Full dataset with weightings and MI (n = 17,132, total sample = 214,755,655) Final respecified model (n = 17,132, total sample = 214,755,655)
ML QML ML QML ML QML ML QML QML
Chi-square test P value <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001
Root mean square error of approximation (RMSEA) (90% CI) 0.062 (0.0590, 0.064) 0.0611 (0.0570, 0.064) 0.066 (0.063, 0.068) 0.0641 (0.060, 0.068) 0.064 (0.062, 0.067) 0.0631 (0.059, 0.067) 0.062 (0.061, 0.064) 0.0621 (0.059, 0.065) 0.024
Standardized root mean square residual (SRMR) 0.035 0.032 0.037 0.035 0.034 0.034 0.032 0.032 0.012
Comparative fit index (CFI) 0.925 0.9271 0.914 0.9161 0.916 0.9191 0.928 0.9301 0.971
Tucker-Lewis Index (TLI) 0.903 0.9061 0.888 0.8921 0.891 0.8951 0.907 0.9091 0.986
1

Denotes robust statistic for QML with Satorra-Bentler correction as produced by the Lavaan package in R. MI, multiple imputation; ML, maximum likelihood; QML, quasimaximum likelihood.