Goodness of fit gof are indicators to tell us whether

This preview shows page 38 - 41 out of 75 pages.

Goodness-of-fit (GOF) are indicators to tell us whether supposed model results at the covariance among observed variable and support relationship amongst them (Hair et al, 2006). There are 3 group need to be reviewed. They are Absolute Fit Indices, Comparative Fit Indices and Parsimony Fit Indices. Absolute fit indices let us know if collected data fits designed model. There are many indices for absolute fit indices. However the most common reported are ratio between Chi-square to degree of freedom, Adjust Goodness-of-fit (AGFA) and Roots Mean Square Error of Approximation (RMSEA). Chi-square was originally considered as the first indicator for absolute fit . Unfortunately, it was too sensitive to sample size. Thus currently, ratio between chi-square and degree of freedom were accepted as alternative. Value of this ratio ranged from 2 to 3 to be considered as acceptable (Schlermelleh-Engel et al, 2003, Wheaton et al. 1977, Joreskog and Sorbom 1993). However it was still sample sensitive then GFI or AGFI which less sample sensitive recommended to be added. GFI/AGFI value greater than 0.85 meant 37
Thesis Report Le Thi Thanh Huyen - Intake 15 good fit model.(Anderson and Gerbig, 1984). RMSEA is another index and also known as “badness of fit index” . RMSEA value lower than 0.8 considered a satisfactory model. (Hu and Bentler 1999, Schlermelleh- Engel et al. 2003) Comparative Fit Indices shows us how well hypothetical model fit compare to null model. Comparative Fit Index (CFI) or Relative Fit Index (RFI) are indicators belong to this group. Acceptable score for CFI or RFI is greater 0.95 (Hu and Bentler 1999) Parsimony Fit Indices help us to select the best model in case there are more than one acceptable model. (Gallagher, Ting, Palmer, 2008). Parsimony Goodness-of fit Index (PGFI) or Parsimony Normed Fit Index (PNFI) are the one to be reported. Value smaller than 0.5 is required for those indices (Muliak, 1989) If GOF indices do not satisfy model fit, we use Modification Indices (MI) and Parameter Change to refine model until ration between Chi-square to degree of freedom drop to acceptable range. Afterward, all others score are going to review before confirm assessment. Seven constructs which measuring by 25 items were specified. Inter-correlation between each pair of them also assigned before running CFA. Initial output revealed unsatisfactory GOF with Chi-square 1,309, degree of freedom 254 and ratio between them was 5.154 . Modification Index (MI) showed us the way to improve model GFI either by assigned relationship among item or removed it. Base on this guidance, we removed items which caused the most significant reduction of Chi-square and arriving at final scale as in Figure 2 with Chi-square 196, degree of freedom 84 which equal to ratio at 2.337. Hoelter was above 200 as illustration in table 4.13 Table 4.13 - Summary Measurement scale Goodness of Fit Original scale Final Scale Critical Value Chi-square 1309 196 df 254 84 Ratio 5.154 2.337 <3 GFI 0.771 0.945 >0.85 RMSEA 0.099 0.056 <0.08 CFI 0.833 0.962 0.95 PNFI 0.679 0.656 0.5 HOELTER 101 254 >200 38
Thesis Report Le Thi Thanh Huyen - Intake 15 SMBG frequency Family support Perceived Benefit HCP interaction

  • Left Quote Icon

    Student Picture

  • Left Quote Icon

    Student Picture

  • Left Quote Icon

    Student Picture