Dear Statalisters,
I want to investigate the mediation relationship with the dependent variable, a latent variable indicated by four binary indicators. Both the independent variable(IV) and the mediator(M) are continuous variables.
Firstly, I use four binary variables to indicate the dependent variable by using Latent Class Analysis. The LCA result shows the dependent variable with 2 classes (binary).
Next, I converted an individual’s posterior class probability to the logistic scale(log odds) and then use it as the dependent variable(logitP0, continuous variable) to run the OLS mediation model.
1. Since the dependent variable is the log odds of class probability from Latent Class analysis, How can I explain the results better as shown below? Can I explain the result that " both IV and M reduces the log odds of probability to be class 0 of Dependent variable; IV is positively associated with M; IV also reduces the log odds of probability to be class 0 of the dependent variable through increasing M. "
2. Is there any better way to run a mediation model after latent class analysis on the dependent variable?
3. If I use class membership (binary variable) as the dependent variable after latent class analysis, what is the stata syntax for the logistic mediation model?
Thank you!
I want to investigate the mediation relationship with the dependent variable, a latent variable indicated by four binary indicators. Both the independent variable(IV) and the mediator(M) are continuous variables.
Firstly, I use four binary variables to indicate the dependent variable by using Latent Class Analysis. The LCA result shows the dependent variable with 2 classes (binary).
Next, I converted an individual’s posterior class probability to the logistic scale(log odds) and then use it as the dependent variable(logitP0, continuous variable) to run the OLS mediation model.
1. Since the dependent variable is the log odds of class probability from Latent Class analysis, How can I explain the results better as shown below? Can I explain the result that " both IV and M reduces the log odds of probability to be class 0 of Dependent variable; IV is positively associated with M; IV also reduces the log odds of probability to be class 0 of the dependent variable through increasing M. "
2. Is there any better way to run a mediation model after latent class analysis on the dependent variable?
3. If I use class membership (binary variable) as the dependent variable after latent class analysis, what is the stata syntax for the logistic mediation model?
Thank you!
