Validating Schizophrenia Symptom Scale

The descriptive analyses, the internal consistency (i.e., Cronbach’s α), and the concurrent validity using Pearson correlation coefficients were analyzed using SPSS 16.0 for Windows (SPSS Inc., Chicago, IL, USA). The confirmatory factor analyses (CFAs), including measurement invariance, were done using LISREL 8.8 for Windows (SSI Inc., Lincolnwood, IL, USA).
Because all the items in the SSS-S were normally distributed (skewness = −0.111 to 0.802; kurtosis = −1.008 to 0.376), a maximum likelihood estimation was used for all CFAs. A second-order model was used for the whole sample and for the separate samples (viz., the sample with schizophrenia, the sample with other mental illnesses, the male sample, and the female sample). The second-order model was also used to evaluate measurement invariance, and the 10 models were as follows:

Model 1M/1G: configural model for mental illnesses/genders;

Model 2M/2G: all first-order factor loadings were invariant between mental illnesses/genders;

Model 3M/3G: all first-order factor loadings and item intercepts were invariant between mental illnesses/genders;

Model 4M/4G: all first- and second-order factor loadings and item intercepts were invariant between mental illnesses/genders;

Model 5M/5G: all first- and second-order factor loadings, item intercepts, and construct means were invariant between mental illnesses/genders;

Fit indices of a nonsignificant χ² statistic, root mean square error of approximation (RMSEA) < 0.08, comparative fit index (CFI) > 0.95, and standardized root mean square residual (SRMR) < 0.08 were used to determine whether the data-fit of the model was satisfactory [24 ,25 ]. Moreover, goodness of fit (GFI), adjusted goodness of fit (AGFI), Akaike’s information criteria (AIC), and consistent Akaike’s information criteria (CAIC) were also reported for the second-order models of four separate samples. A nonsignificant χ² statistic was also used to test measurement invariance. In addition, ΔRMSEA and ΔCFI < 0.01 suggest that factor loadings, item intercepts, and construct means were invariant across measurements. ΔSRMRs < 0.03 and < 0.01 also suggest that factor loadings and item intercepts were invariant [17 (link),26 –29 ].