1
$\begingroup$

Vandenberg & Lance (2000) explain that measurement invariance in SEM is important because the variance-covariance matrix of a group, $\Sigma$, and item scores, $X$, can be decomposed, such that

$$X_k^g=\tau_k^g+\Lambda_k^g\xi^g+\delta_k^g$$

and

$$\Sigma^g=\Lambda_X^g\Phi^g\Lambda_X'^{g}+\Theta_\delta^g$$

where $g$ is a group identifier, $k$ is a variable identifier, $\tau$ is an intercept, $\lambda$ is a regression slope relating $X$ to $\xi$, $\delta$ are unique factors, $\Lambda$ is an item's factor loadings on the latent factor $\xi$, $\Phi$ is the variance-covariance among $\xi$, and $\Theta$ is the unique variance to an item.

Is my diagram below correct in representing this relationship? I expect the $\delta$ and $\theta$ parts are incorrect, so I am unsure of how to diagram them.

enter image description here

$\endgroup$

1 Answer 1

0
$\begingroup$

You've drawn the diagram a little unconventionally, for example you've only drawn one indicator, and $\Phi$ is often irrelevant when you're looking at invariance, so it is not modelled. $\theta$ is usually thought of as being equal to 1, and so can also be ignored / removed.

Apart from that, looks OK to me.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.