I think your reasoning is very balanced, especially how you combine statistical thresholds with theoretical considerations. I agree with you that relying too strictly on ...
I think your reasoning is very balanced, especially how you combine statistical thresholds with theoretical considerations. I agree with you that relying too strictly on cutoffs like CFI ≥ .95 can sometimes be unrealistic, particularly with more complex models or smaller samples.
That said, I’m a bit curious about how far is “acceptable” for you in practice. For example, your CFI = .94 is very close to .95, so it feels reasonable to accept—but would you still accept the model if it dropped to, say, .92 or .90? Where would you personally draw the line?
Also, I like that you mentioned checking modification indices, but do you think there’s a risk of overfitting the model if we rely too much on them? Maybe it would be helpful to also report other indices like SRMR or χ²/df to give a more complete picture of model fit.
Overall, I agree with your point that model evaluation should be flexible, but I think it’s also important to be transparent about how much deviation from the “ideal” we are willing to accept and why.
That said, I’m a bit curious about how far is “acceptable” for you in practice. For example, your CFI = .94 is very close to .95, so it feels reasonable to accept—but would you still accept the model if it dropped to, say, .92 or .90? Where would you personally draw the line?
Also, I like that you mentioned checking modification indices, but do you think there’s a risk of overfitting the model if we rely too much on them? Maybe it would be helpful to also report other indices like SRMR or χ²/df to give a more complete picture of model fit.
Overall, I agree with your point that model evaluation should be flexible, but I think it’s also important to be transparent about how much deviation from the “ideal” we are willing to accept and why.