Alice Geng posted on Thursday, March 27, 2014 - 11:35 am
The aim of my study is to describe the trajectory of AC over 2 years using latent growth model (LGM).
AC is a dichotomized variable (0=poor AC status, 1=good AC status), and AC is measured 4 times over 2 years for the subjects, however, each measurement of AC was assessed at the different time for the participants. To be more specific, I have one month baseline window for each patient to report their AC status in the first year, and during the follow-up 3 months, the AC status was recorded only if the second AC was different from the baseline AC. And the same procedure was followed in the second year. Therefore, each AC was assessed at different time for each participant.
I am not sure how to deal with the time intervals and assign factor loadings in the latent growth model. If the factor loading for the first AC is assigned to 0 and the fourth factor loading for the last AC is assigned to 1, can I use the proportions of people with poor AC status as the factor loadings for the second and third AC measures?
Additionally, do you know if there are any articles discussing dichotomized variables in the latent growth models?
Thank you very much for taking the time to answer my questions.
We discuss Mplus growth modeling with binary outcomes in our Topic 3 and 4 handouts and videos on our website. This includes modeling with individually-varying times of observation.
Your outcomes, however, are different from the typical binary outcomes of growth modeling in that you require an AC status change - it sounds like you can only have specific sequences of 0/1 and that limitation needs to be reflected in the modeling.