I have collected the responses from 584 respondents, SEM model is fine but the GFI is showing exactly 1, I think it should be less than 1 and greater than 0.7 according to the threshold set. Please guide me in this case. Is there any minor changes to be made in the values of the respondents(Likert Scale.. Degree of superlative, ranging from 1 to 5) which are biased in nature.
The GFI is actually an outdated fit index for which a value of 1 is supposed to indicate good model fit. However, the index looks good for most models (even ones that in actuality don't fit very well) and is not used much anymore. The most rigorous SEM test of fit is provided by the chi-square test of model fit.
GFI 1 means yours model is good fit. The level of acceptance of GFI more than 0.90. Please refer to support your results through this literature
(Bentler and Bonett, 1980; Hooper et al., 2008; Hair et al., 2010; Kline,
2011).
Hello Shivasharana,
If your SEM model is saturated (no degrees of freedom), then any fit index is meaningless, since the model will perfectly reproduce the observed relationships in the data set. I couldn't tell from your query whether this might be what you're reporting.
I do agree with the prior post that GFI isn't a particularly good index on which to rely. Have a look at David Kenny's very readable summation of "the usual suspects" among fit indices for SEM: http://davidakenny.net/cm/fit.htm
Good luck with your work.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Structural Equation Modeling