LTA and odds ratio question PreviousNext
Mplus Discussion > Latent Variable Mixture Modeling >
Message/Author
 Jessica Larsen posted on Wednesday, May 08, 2013 - 11:08 am
I am conducting a Latent Transition Analysis (LTA) in Mplus and using a covariate to examine the interaction of the covariate on the transition patterns. My sample is a clinical sample, and I am examining transitions over the course of treatment. My covariates represent categories with low base-rates (i.e. males who report past sexual abuse). What I am finding, is that when the covariate is included in my model, certain transition paths (ie c1#2 to c2#3) are estimated to have 0 probabilities. This is of substantive interest and makes sense theoretically. However, statistically, this zero estimate causes one of my estimated odds ratios to be very large (i.e. 57725). Is this accurate? How can I report this in my write-up?
 Linda K. Muthen posted on Thursday, May 09, 2013 - 9:47 am
I would have an asterisk in the table and say something like:

Odds rations cannot be computed because of a perfect correlation.
 JuliaSchmid posted on Wednesday, February 06, 2019 - 12:20 am
Hi Linda

I run a LTA and I would like to report Odds instead of Transitional probabilities. Do I have to calculate them (CI included) by hand or is there a possibility that Mplus give it in the output?

Thank you for your help,
Julia
 Bengt O. Muthen posted on Wednesday, February 06, 2019 - 5:22 pm
You can express them in Model Constraint using parameter labels given in the Model command. Then you get CIs too.
 JuliaSchmid posted on Thursday, February 07, 2019 - 7:44 am
Hi Bengt!

I would appreciate, if you would explain the procedure in more detail: How can I calculate the Odds Ratio exactly via Model Constraint-Command?

Let's say I run a LTA with two measurement points and 4 classes each. An my class indicators/variables are:

Steuko_1 Befreg_1 Selko_1
Steuko_2 Befreg_2 Selko_2

Thanks in advance!
 Bengt O. Muthen posted on Thursday, February 07, 2019 - 5:28 pm
I assume your variables are binary.

See our Short Course Topic 6, slides 48 and on.
 JuliaSchmid posted on Friday, February 08, 2019 - 5:50 am
Perfect, thanks to the sliedes, I understand now the logic of the input.
No, my variables are contiuous - I run a LPTA. Does this fact change the procedure mentioned on slides 48f?
 Bengt O. Muthen posted on Friday, February 08, 2019 - 7:04 am
Not if the odds ratios you mention refer to only the transitions, not the outcomes. See web note 13 which describes such odds ratios:

Muthén, B. & Asparouhov, T. (2011). LTA in Mplus: Transition probabilities influenced by covariates. Mplus Web Notes: No. 13. July 27, 2011.
 Marina Epstein posted on Monday, May 04, 2020 - 10:56 am
I am running a simple Markov model with specified thresholds for the indicator variables (below). I get predicted latent probabilities of moving form one state to another. Is there a way to be able to test whether the probability of moving from state a to state b is more likely than moving from state a to state c, etc?

MODEL:
%OVERALL%
C2 ON C1;

MODEL C1:

%C1#1%
[smk21$1@15];
[vap21$1@15] ;

%C1#2%
[smk21$1@-15] ;
[vap21$1@15] ;

%C1#3%
[smk21$1@15] ;
[vap21$1@-15] ;

%C1#4%
[smk21$1@-15] ;
[vap21$1@-15] ;

MODEL C2:

%C2#1%
[smk23$1@15];
[vap23$1@15] ;

%C2#2%
[smk23$1@-15] ;
[vap23$1@15] ;

%C2#3%
[smk23$1@15] ;
[vap23$1@-15] ;

%C2#4%
[smk23$1@-15] ;
[vap23$1@-15] ;
 Tihomir Asparouhov posted on Tuesday, May 05, 2020 - 10:35 am
Yes.

ANALYSIS: parameterization=probability;

model:
%overall%
C2#1 on C1#1 (b);
C2#2 on C1#1 (c);

model test: 0=b-c;

and divide the p-value by 2 to get a one-sided test.

https://en.wikipedia.org/wiki/One-_and_two-tailed_tests
 Marina Epstein posted on Wednesday, May 06, 2020 - 8:52 am
Thank you! I want to make sure I am using the right set up for my Markov model. I am modeling transition probabilities between four behaviors (smoking, vaping, both, neither) at two time points.

My other question is whether there is information in the TECHs that would provide CIs or other tests that I should be reporting from this model.
 Bengt O. Muthen posted on Wednesday, May 06, 2020 - 2:21 pm
Have a look at Mplus Web Note 13 on our website. It shows several uses of Model Constraint to look at odds ratios.
 Bengt O. Muthen posted on Wednesday, May 06, 2020 - 4:41 pm
Have a look at Mplus Web Note 13 on our website. It shows several uses of Model Constraint to look at odds ratios.
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: