AMOS SEM model drawing rules

[kissgigo]

hi!

This is my first attempt to draw a model in SPSS AMOS 19. I am confused by seeing that I cannot draw covariance arrow between two variables, A and B provided there is a path arrow pointing from C to one of them.

Could you explain it to me?

What is the difference between drawing two separate path arrows from A to B and vica versa and drawing a covariance arrow between them?

I know these questions sound elementary, but I could not figure out these from the Help and even from the User Guide.. :-(

cheers

[CowboyBear]

hi!

This is my first attempt to draw a model in SPSS AMOS 19. I am confused by seeing that I cannot draw covariance arrow between two variables, A and B provided there is a path arrow pointing from C to one of them.

It's hard to answer this without seeing the path diagram (attaching it might be helpful). At a guess I'd say it might be a problem with the covariance term forcing the model to be obviously unidentified.

If you're just getting started with Amos you might find Barbara Byrne's book on SEM with Amos helpful - see if you can get hold of a copy!

What is the difference between drawing two separate path arrows from A to B and vica versa and drawing a covariance arrow between them?

Conceptually speaking, two path arrows would mean you believe that each of the two variables has a causal effect on the other. A covariance term simply means you're allowing the variables to be correlated without making specific statements about their causal relationship. In practical terms I'm not sure if you'd be able to estimate a model with bidirectional causality (path arrows in both directions) between two latents - I can't see how such a model could be identified.

[kissgigo]

Thank you very much, it was very helpful.

[arindam]

I have a similar problem and need help.

The model I am defininng has both endogeneous latent variables and exogeneous latent variables. I would like to define causal relation (one-directional) from exogeneous to endogeneous variables and at the same time covariance relationship between endogeneous variables. But I am not able to do both. If I draw unidirectional lines from A (exogenoeus latent) to B (endogeneous latent) and C (exogenoeus latent) to D (endogeneous latent) then I am not able to draw a birectional relationship between B and D. On the other hand, if I draw bidirectional lines first then I am not able to draw the unidirectional lines. Am I making a mistake? Or is my model definition wrong?

Please advise.

[Lazar]

So the model is:

B<-------->D
^          ^
|          |
|          |
|          |
|          |
a          c

If so yes you can run this model but you should be aware that your correlation between B and D will be the correlation in the residuals of those variables.

[arindam]

Yes, that's how I want to specify it. My problem is that AMOS does not allow me to draw the lines from a to B and c to D once I draw the bidirectional line between B and D.

[Lazar]

Odd. You will have to give more detail. Does AMOS give some kind of syntax? Can you run a CFA with just correlations between a b c and d?

I have generated some data based on the model I drew above (find attached). See it AMOS will let you run the model with this data.