Reliability and validity test
The research conducted a reliability and validity analysis of the questionnaire in the study, using the Cronbach’s Alpha coefficient to verify the reliability of the questionnaire and questions on SPSS25.0 software, and then using factor analysis to verify the content validity of the questionnaire tools and questions. Researchers often use the Cronbach’s alpha coefficient to examine the internal consistency of measurement tools (Kaplan, 2009), in consequence of the Cronbach’s alpha was above 0.7, means the tools are acceptable, 0.8 means good reliability, and 0.9 means preferable (Kline, 2011; George & Mallery, 2003). Thus, if the reliability of the questionnaire item is greater than 0.8, it can be judged that the questionnaire tool has high reliability and stability. If the reliability of the questionnaire item is greater than 0.7, it can be judged that the reliability of the questionnaire is average and within the acceptable range.
Through internal consistency analysis, the Cronbach’s Alpha coefficient of the web-based learning tool scale (13 items in WBLTs) is 0.954>0.9, which is highly reliable. The Chinese version of the general self-efficacy scale (10 items in GSES-C) has Cronbach’s The Alpha coefficient is 0.939>0.9, which has a high reliability. The Cronbach’s Alpha coefficient of the depression-anxiety-stress scale (21 items of DASS) is 0.959>0.9, which has high reliability. The Cronbach’s Alpha coefficient of the family quality of life scale (FQLS) (25 items) is 0.961>0.9, which has high reliability. the Cronbach’s Alpha coefficient of the general health questionnaire (12 items of GHQ) is 0.702>0.7, and its reliability was within an acceptable range.
Also, factor analysis and Kaiser-Meyer-Olkin (KMO) test were always used to confirm the construct validity, and the content validity was examined by literature review and expert appraisal (Catalano, 2018). This study conducted an exploratory factor analysis on five questionnaires to test the structural validity of the questionnaire, selecting the maximum variance orthogonal rotation method, KMO and Bartletts’ test, and principal component analysis (extract factor if the characteristic value is greater than 1, the absolute value of the factor load is greater than 0.5, and the results are shown in the table). Utilizing factor analysis, the KMO value of the web-based learning tool scale is 0.940>0.9 (p<0.000, satisfied with the Bartletts’ test), the factor loading is between 0.647~0.848, and the cumulative variance contribution rate of three common factors is 79.046%. Its validity is relatively high. The KMO value of the Chinese version of the self-efficacy scale (10 items of GSES-C) is 0.928>0.9 (p<0.000, satisfied with the Bartletts’ test), the factor loading is 0.487~0.795. The cumulative variance contribution rate of the co-factors is 73.502%, and its validity is very high. The KMO value of the depression, anxiety and stress scale (21 items of DASS) is 0.947>0.9 (p<0.000, satisfied with the Bartletts’ test, the factor loading is 0.625~0.809, and the cumulative variance contribution rate of 3 common factors is 73.244%, and its validity is high. The KMO value of the Family Quality of Life Scale (FQLS 25 items) is 0.950>0.9 (p <0.000, satisfied with the Bartletts’ test), the factor loading is 0.506~0.934, the cumulative variance contribution rate of 5 common factors is 79.334%, and its validity is high. The KMO value of general health scale (12 items on GHQ) is 0.917>0.9 (p<0.000, satisfied with the Bartletts’ test), the factor loading is 0.660~0.832, and the cumulative variance contribution rate of extracting 2 common factors is 75.688%, and its validity is very high.
Finally, confirmatory factor analysis (CFA) was used to verify the reliability, convergence validity, and discriminate validity of each questionnaire. As shown in Table 3 and Table 4, four major measurement tools as follows: web-based learning tool scale, depression-anxiety-stress scale, general self-efficacy scale, and family life quality scale, are mainly used in confirmatory factor analysis. Firstly, after the test of the proposed model, the fitting index of the proposed model was very unsatisfactory. Therefore, the independent variable of the general health questionnaire was directly removed. Furthermore, during the process of confirmatory factor analysis and structural equation model construction, the general health questionnaire was excluded from the SEM. In addition, it can be seen from Table 3 and Table 4 that the reliability and validity of other questionnaires reached a good level excluding the general health questionnaire.
Table 3 Reliability and Convergence validity