Analyzing panel data with 3 time points: latent growth/curve models

#1
Hello,
I have panel data containing 3 time points.
Some of the papers I read used latent curve models. I'm trying to employ this idea as well but there were some bumps in the road.
First: I'm using AMOS, but it seems that this program is only equipped to deal with one latent variable that changes over time, while in my data I have measurements of both time- varied DV and time- varied covariates. Am I missing something?
Second, since I don't have Mplus, is there another solution ? R? Something less "fancy" than latent curve modeling? My main finding (when I use simple regression) is that : DV t3 is predicted by DV t1 IV t1 and the interaction between them (iv t1 and dv t1). I'm not sure but my intuition is that it is very similar to latent growth, or am I wrong?
Thanks
 

Lazar

Phineas Packard
#2
Lavaan in R has a specific growth curve function. I have also coded up growth curve models in lavaan by hand with multiple constructs of interest and see no reason why you could not include time varying covariates. Of course you could drop a focus on latent growth curves entirely and instead using multilevel modeling to estimate growth curves. I believe if you search this forum a user by the name of spunky has coded up example in R using latent growth curves and multilevel approaches.
 

Lazar

Phineas Packard
#3
While I am at it I see no reason why you cannot do what you want to do in AMOS. I have never used the program myself but provided you can constrain parameters and AMOS provides estimates the mean structure you should be fine.
 
#5
Thanks this is all very helpful
I do have one more theoretical question - is the interaction between the dv and iv in t1 similar to any effects the iv has on slope of the dv?
 
#6
While I am at it I see no reason why you cannot do what you want to do in AMOS. I have never used the program myself but provided you can constrain parameters and AMOS provides estimates the mean structure you should be fine.
I created a model in which the dv is latent using AMOS and then manually calculate intercept and slope to the iv , not very elegant or accurate I think.