Example 8.14 with more than two time ...
Message/Author
 João Miguel Picoito posted on Monday, November 04, 2019 - 2:08 am
I want to use syntax from example 8.14 for a latent transition analysis model with a covariate influencing transition probabilities, but with five time points.
I don't know if the %overall% part of the syntax is correct.

For a time-invariant covariate (x):
MODEL:
%OVERALL%
C2 ON C1; C3 ON C2; C4 ON C3; C5 ON C4;
C1 ON x; ! (is this part correct? or should I add C2 on x; C3 on x; and so on?)
MODEL C1:
%C1#1%
C2 on x;
[u11\$1] (1);
…….
MODEL C2:
%C2#1%
C3 on x;
[u21\$1] (1);
……..
MODEL C3:
%C3#1%
C4 on x;
[u31\$1] (1);
……….
MODEL C4:
%C4#1%
C5 on x;
[u41\$1] (1);
……..
MODEL C5:
%C5#1%
[u51\$1] (1);
…….

And for a time-varying covariate (x1, x2, x3, x4, x5)? Is the below syntax correct?

MODEL:
%OVERALL%
C2 ON C1; C3 ON C2; C4 ON C3; C5 ON C4;
C1 ON x1 x2 x3 x4 x5;
MODEL C1:
%C1#1%
C2 on x1;
[u11\$1] (1);

MODEL C2:
%C2#1%
C3 on x2;
[u21\$1] (1);
…….. (and so on).

Thank you!
 Bengt O. Muthen posted on Monday, November 04, 2019 - 9:37 am
You are specifying a very complex LTA that will be computationally heavy. In principle there is no error in the syntax as far as I can see. Although I don't understand why x2-x5 would influence C1 if they happen later.

However, I think you should study the new and better way to do LTA described here:

http://www.statmodel.com/RI-LTA.shtml

Note also that the forthcoming version 8.4 will be much faster for LTA.