| Message/Author |
|
|
| Jan Ivanouw posted on Monday, November 19, 2018 - 12:27 pm
|
|
|
Hi,
I have 8 continuous indicators measured twice for the same persons.
At each time point a latent class analysis with 3 classes fits well.
I would like to make a transition analysis showing the transitions between the three classes for the persons.
As an attempt, I had the two latent class models calculate the best fitting class for each person and made a 3x3 table with these.
However, I am sure there must be a better way, maybe a comprehensive model which makes the latent class analysis and the transition analysis in a single process. A common LTA-model will not do, since it needs categorical indicators.
Any advice?
Best, |
|
|
|
|
| LTA handles categorical indicators also - see UG ex 8.13 and the examples in the LTA paper section. |
|
|
| Jan Ivanouw posted on Tuesday, November 20, 2018 - 5:01 am
|
|
|
| The problem is that my indicators are continuous, and I have not found any Mplus syntax which seems to work in this situation. |
|
|
|
|
Going from categorical outcomes to continuous, you just change the invariance statements to concern intercepts in brackets instead of thresholds in brackets, so e.g.
time 1:
[y11-y15] (p1-p5);
time 2:
[y21-y25] (p1-p5); |
|
|
| Jan Ivanouw posted on Wednesday, November 21, 2018 - 3:14 am
|
|
|
| So simple! Thank you very much |
|
|
| Jan Ivanouw posted on Friday, November 23, 2018 - 6:34 am
|
|
|
I am performing a latent transition analysis with Mplus version 8.2, and I don’t understand why I am getting the following error message:
“*** ERROR
The following variable has different values for different repetition values within the same ID.
Specify the variable in the LONG option in the DATA LONGTOWIDE command to convert it from long to wide format or exclude it from the analysis by using the USEVARIABLES option in the VARIABLE command.
Problem with variable _TV1 for ID 3013
There is no variable named _TV1 in the data set.
I have sent my license number, output and data files (JanIvanouwLTA.ZIP) to support@statmodel.com. |
|
|
| Back to top |
