Hi dears,
my structural model assessment results are the same with CFA results (measurement model assessment), including TLI, SRMR, CFI, RMSEA, chi-square, etc. do you know what's the reason?
(CR, AVE, Collinearity test results are all good)
many thanks!!
Perhaps the models are equivalent (imply the same covariance structure). Without knowing more about both the models, it is impossible to tell.
Christian Geiser
Hi Christian,
my model has three latent variables, and each of them has four indicators. my hypothese are:
H1: A has an positive impact on B
H2: B has an positive impact on C
H3: A has an positive impact on C
(i think it's a simple model)
i have searched some related questions on some forums, do you think is it possible it's a saturated model? (but the chi-square of my model is not zero). - actually i' m not familiar with 'saturated model' maybe i'm totally wrong....
Hi Rowenga,
as Christian has said. You test a partial mediation model without any restriction (fixed-to-zero parameters) on the latent level. Hence it is statistically equivalent to the CFA model, a regression model and dozens of other models.
If you really want to test mediation (and differentiate it from many other models) perhaps you have some canditates for instruments (for X and / or M) in your data set.
Kline, R. B. (2015). The mediation myth. Basic and Applied Social Psychology, 37(4), 202-213. https://doi.org/10.1080/01973533.2015.1049349
Thoemmes, F. (2015). Reversing arrows in mediation models does not distinguish plausible models. Basic and Applied Social Psychology, 37(4), 226-234. https://doi.org/10.1080/01973533.2015.1049351
HTH
Holger
The models are saturated at the latent (structural) level. You have three latent variables (factors) A, B, C. There are three covariances between them. In the CFA, you simply estimate the three covariances. In the SEM (latent path model), you use the three covariances to estimate three paths between the latent variables. Therefore, the CFA and SEM in this case have the same df and chi-square fit.
The model is (apparently) non-saturated in the measurement part (the portion that connects the indicators (observed variables) to the latent variables (factors). That's why you get df > 0 and chi-square > 0.
Holger Steinmetz
Hi Holger,
thank you for your response! i've tried restrict the parameters (i choose three paths for all three latent variables: from latent variable to indicator, and restrict those paths to 1), but still get same results, do you know the reason? many thanks
Christian Geiser Hi Christian,
thanks for your response! i' ve tried to restrict the parameters but still get same results (as i answered Holger), could you please suggest some possibility for this? thanks!
- Badges
- Science topic
- Similar topics
- Linguistics
- Composition Studies
- Publication