freek bucx posted on Saturday, October 21, 2006 - 4:59 am
I have a question. I have tested a SEM model. Fit indices reveals that the chi square is quite high: 138.86, df=47, p <.00. Other fit measures, however, show an acceptable fit: RMSEA = 0.042; Standardized RMR = 0.025; AGFI = 0.95; CFI = 0.99. Can I accept this model, although the chi square is rather high (N = 1,050).
One important piece of information is what the sample size is and the scale of the dependent variables. If the sample size is very large, you might be able to say that chi-square is too sensitive. You can do a sensitivity analysis by freeing parameters in the model until you get an acceptable chi-square. Then compare the parameters in your original model to the same parameters in the new model. If the values are close, then you might be able to claim that chi-square was senstive because of a large sample size.
freek bucx posted on Monday, October 23, 2006 - 10:29 am
Thank you for your quick answer! Such a sensitivity analysis sounds like a good idea!
Another question: have you good some reference to a book or article in which this use of the sensitivity analysis is explained further?