Growth terms with/out a time-varying ...
Message/Author
 YUN HWAN KIM posted on Wednesday, September 26, 2018 - 12:43 pm
Dear Muthen(s)

Please let me ask if I can interpret the results as below.

In the growth model of "A", both linear and quadratic terms were significant. After including "B" as a time-varying covariate, the linear term became much smaller in its magnitude (about two-thirds disappeared), and the quadratic term was no longer significant.

And I did the same after switching "A" and "B". In the growth model of "B", both linear and quadratic terms were significant. After including "A" as a time-varying covariate, the linear term became bit smaller in its magnitude (about one-fourths disappeared), and the quadratic term was still significant.

Based on the above, can I interpret (or infer) that the effects of "B" on the growth of "A" seem to be greater than the effects of "A" on the growth of "B"?

 Bengt O. Muthen posted on Wednesday, September 26, 2018 - 6:26 pm
Your time-varying covariates probably have a trend. You can explore that by growth modeling.
 YUN HWAN KIM posted on Wednesday, September 26, 2018 - 9:19 pm
Dear Muthen,

I appreciate your quick answer. And your guess is correct. Both "A" and "B" presented similar curvilinear trends over time. And they are theoretically expected to have a mutual influence.

In order further to figure out the relations between the two (one of them was the above question: if the effect of "A" on "B" is stronger than that of "B" on "A"), I would ideally run a parallel process growth model, but it comes with numerous estimation problems.

My last resort was, therefore, to run a growth model of "A" while including "B" as a time-varying covariate and to run a growth model of "B" while including "A" as a time-varying covariate (and I obtained the above-mentioned results). Seeing that "B" diminished the growth pattern of "A" to a greater extent than "A" diminished the growth pattern of "B", I reasoned that this may indicate the stronger effects of "B" on the growth "A" than the other way round. But I am wondering if this reasoning makes sense (although I am aware that I did not technically test it), or if it is too much stretched-out interpretation/inference.
 Bengt O. Muthen posted on Friday, September 28, 2018 - 11:07 am
This general question is suitable for SEMNET.
 YUN HWAN KIM posted on Friday, September 28, 2018 - 11:06 pm
Dear Muthen