second-order latent growth curve model

[Sara]

Hello everyone,

I am estimating a second-order latent growth curve model with lavaan. At each survey wave depression is specified as a latent construct measured by five ordinal items; the items are included in the ordered = argument so that lavaan treats them as ordinal.

In the model the wave-specific latent factors are regressed on lagged time-varying covariates to test whether these covariates predict subsequent levels of depression. Although the covariates themselves are ordinal  I have not included them in ordered =, because they are used as external predictors.

This works as expected, but now I would also like to estimate the relationships within the same wave, not only the lagged effects.

My questions:

• Is it appropriate simply to add regressions (e.g. F2002 ~ drug_2002 + health_2002) alongside the lagged effects?

• Or would it be better practice to estimate a separate model where I replace the lagged effects with the same-year effects?

Any advice would be greatly appreciated.

Thank you!