Equating transitional probabilities PreviousNext
Mplus Discussion > Latent Variable Mixture Modeling >
Message/Author
 Irina Bondarenko posted on Friday, August 28, 2020 - 9:48 am
I run LTA analysis on 4 waves of survey data. I use 9 3-level variables and specify model with 6 classes. I constrained item-response probs and test invariance of transitional probabilities between the waves. Here is my code.
!Time 1 to Time 2
C2#1 on C1#1 (1001);
......
c2#5 on C1#6 (5006);
! Time 2 to Time 3
c3#1 on c2#1 (1001);
-------
c3#5 on c2#6 (5006);
!Time 4 to Time 3
c4#1 on c3#1 (1001);
----
c4#5 on c3#6 (5006);
With this model I estimate 153 parameters.However, examples that I found on the web (see https://www.statmodel.com/download/LTA_DP_FINAL.pdf) call for additional constrain
[C2#1] (11111);

.....
[C2#5] (55555);
[C3#1] (11111);
.....
[C3#5] (55555);

[C4#1] (11111);
......
[C4#5] (55555);
It doesn't seem right unless unless I'd like to constrain prevalence of LC. Can you please clarify this.
 Bengt O. Muthen posted on Saturday, August 29, 2020 - 4:43 pm
Have a look at Chapter 14 of the UG, pages 558-559. The table on page 559 shows that both a and b contribute to the logits in the cells and therefore the transition probabilities. The a's don't correspond to the prevalence of the LC, but are intercepts.
 Irina Bondarenko posted on Monday, August 31, 2020 - 12:33 pm
Dear Dr.Muthen,
Thank you for pointing out the differences in parametrization. After Reading Chapter 14 it's clear to me.
Back to top
Add Your Message Here
Post:
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Password:
Options: Enable HTML code in message
Automatically activate URLs in message
Action: