|
|
 |
| DV latent variable with continuous an... |
 
|
| Message/Author |
|
|
| Eric M. posted on Thursday, December 14, 2017 - 7:32 pm
|
|
|
Hi. I am considering using a latent variable (with 3 indicators) as a DV of an SEM model (with several predictors). Two of the indicators are continuous. The third indicator is categorical (either binary or ordered 3 category). I assume that I should declare the categorical indicator as such. (1)Would the latent variable be considered a continuous or categorical variable?
It is my understanding based on reading your website and manual “When at least one factor indicator or other observed dependent variable is binary or ordered categorical, Mplus has seven estimator choices: WLS, WLSM, WLSMV, ML, MLR, MLF, and ULS.” (2) What guidance (e.g., articles, recommendation) is there for choosing the appropriate estimator?
Last, the reason I may use categorical variable as an indicator in my latent DV, is because the raw data is a little problematic (non-normal). I have log transformed the variable, which helps. But it led me to wonder (3) should used one of the robust estimators that does require normality? How are fit indices interpreted when using these alternative like MLR or MLM. The output is a little different if I am remembering correctly. |
|
|
|
|
(1) Continuous. Using BY implies a continuous latent variable. See also our Topic 2 short course video and handout.
(2) See our FAQ:
Estimator choices with categorical outcomes
(3) Use MLR for which chi-square and other fit indices are judged as usual. More advanced - declare the variable as censored or two-part; see our book "Regression and Mediation Analysis using Mplus". |
|
|
| Back to top |
|
|
