Interpretating SEM estimates for 2 different datasets

Question

I have a fitted a SEM model to World Values Survey (WVS) dataset. I regressed three latent factors (MnP, RnF and V) standing for "Money & Property" "Relationships & Family" and "Violence" on an exogenous latent variable : "Religion". Here is a figure of my model :

Now, when I apply this model to the Netherlands, I get this regression table :

And when applying it to Malaysia subset, here is what I get :

I would like to know if I understand correctly these tables. In the case of Malaysia, all three linear regressions are significant but does the lower size of the estimate and the high dimension of the dataset makes these relations spurious ? Also, the estimator used here is maximum likelihood instead of OLS, does it change something to my interpretation ?

EDIT : Since my question is related to an assignment of my univeristy, I wish to get no more detailed answers than the one I already received because I don't want to lack scientific integrity.

Answer 1

Small things first:

I'm not sure what you mean by the high dimension of the dataset mattering.

OLS is ML, so the fact that you used ML instead of OLS doesn't matter.

The structural part of your model is saturated. You could replace those three regressions with covariances, and it won't really change anything.

Big thing:

You are comparing the relationships of latent variables, but you do not appear to have considered whether these latent variables are actually equivalent. If the loadings of MnP, RnF and V are not the same across the two countries, then you are not looking at regressions of the same things. You need to check for invariance of the model across groups (the measurement model) before you look at the structural model. You do this by fitting a multiple groups model.