1. What must be the nature of a covariate when doing ESEM (on ordinal outcomes): can it be continuous, ordinal, binary?
2. As covariate cannot be specified in CATEGORICAL, how does Mplus 'know' the distribution of the covariate? Could you please give some references where this topic is explained?
3. You wrote in another thread: "You can include the covariates in the analysis by mentioning their variances in the MODEL command. They are then treated as dependent variables and distributional assumptions are made about them". Could you please give references where such a case is shown?
Thanks Linda for your answers. I am realizing that I am still on a steep learning curve, here...
3. Sorry to insist, but could you please specify where in the Mplus literature I could find explanation on the matter (I looked in the user guide but couldn't find any examples)? I'm not sure I really understand the reasons and implications of changing an observed exogenous variable from independent to dependent, hence I'm not sure it is what I want to do.
I am hoping that you can assist with some confusion that I have regarding covariate modelling.
Briefly, I am conducting a multi-mediation analysis. I have incorporated several standard observed demographic covariates (age, gender, race) as well as a latent socioeconomic status variable in a model that uses both latent and observed IVs and an ordered categorical DV (using WLSMV estimation).
In the model statement, if I use the statement, SES on AGE RACE MALE, I receive a warning message indicating that two dichotomous variables (RACE and Male) were declared as continuous. It appears that means and variances are estimated for these variables.
However, if I attempt to correct this by switching to covariances (SES with AGE RACE MALE) I receive a message indicating the following: WARNING: THE RESIDUAL COVARIANCE MATRIX (THETA) IS NOT POSITIVE DEFINITE. THIS COULD INDICATE A NEGATIVE VARIANCE/RESIDUAL VARIANCE FOR AN OBSERVED VARIABLE.....
Removing either of these statements entirely is not an option as model fit is bad and modification indices suggest otherwise.
Might you have any suggestions as to which is correct? Particularly when both options provide warning messages
A. Ardèvol posted on Monday, January 13, 2020 - 11:27 am
Dear Drs. Muthén:
If I mention the variance of my exogenous variables in the MODEL command, the output shows estimates for the relationship between these exogenous variables (WITH's), without using SAMPSTAT. For example:
X1 WITH X2 0.051 0.015 3.359 0.001
Can I understand these estimates as estimates of the correlation among my exogenous variables? Is it a correlation or a covariance instead?