I did a factor analysis and identified two factors (i.e. intrinsic satisfaction and extrinsic satisfaction). Then on Mplus I did a regression of the outcome variable on factor scores: overall satisfaction on intrinsic factor score, overall satisfaction on extrinsic factor score. Below is the result.
Model fit is good.
Regression coefficient of overall satisfaction on extrinsic factor score is very low. Is this problematic please (seems that only one factor instead of two factors are driving overall satisfaction)? Could I report this in publications? Thank you!
Hello Ning Yan,
if you did already a factor analysis, why don't you directly incorporate a measurement model in the overall model? Besides that, your structural model is saturated, hence there is a "perfect fit".
Best,
Holger
In addition to Holger Steinmetz 's excellent comments, what you are seeing is likely a case of redundancy. Intrinsic and extrinsic satisfaction are strongly correlated (.773). There seems to be no unique contribution of extrinsic satisfaction above and beyond intrinsic satisfaction when predicting overall satisfaction. I don't think that this finding is per se "problematic" unless there are other issues with your analysis.
Thank you both for your answers! Holger Steinmetz Christian Geiser
The full model does not show a good model fit (plz see below). Also, intrinsic has no significant impact on overall satisfaction in this model.
Could you help look at the model fit and diagram please, and any advice to improve it? Thank you so much!
Hi, yes indeed.
There is lots of work to do in the proper measurement model.
I would start from the start, think about how you conceptualize your both constructs and then select the 2-3 best indicators. Alternatively try to locate the diverse latent structure underneath the indicators that you have.
Best,
Holger
Hi again,
if your latent variable correlates with other variables, 2 indicators are fine and all parameters (loadings, error vars, and the latent's variance) can be estimated.
If you not, you could fix both loadings and save one df. This, however, would rest on the viability of this equality-assumption.
Perhaps, this paper can help where I could demonstrate how the handling of a misspecified model could work
Rosman, T., Kerwer, M., Steinmetz, H., Chasiotis, A., Wedderhoff, O., Betsch, C., & Bosnjak, M. (2021). Will COVID‐19‐related economic worries superimpose health worries, reducing nonpharmaceutical intervention acceptance in Germany? A prospective pre‐registered study. International Journal of Psychology, 56(4), 607-622.
Article Will COVID-19-related economic worries superimpose health wo...
Hope this helps
Holger
- Badges
- Science topic
- Similar topics
- Chemistry, Organic
- Organic