Random effect without a lagged variable PreviousNext
Mplus Discussion > Dynamic Structural Equation Modeling >
 Julian Rubel license number: SABC60080610 posted on Tuesday, November 14, 2017 - 8:12 am

I have a problem with my syntax and I hope you can help me. I have a model where I want to check the mediation from alliance (Inter) to symptoms (HSCL) over Coping (Intra) and the other way round over two sessions(L1 means that this is a manually lagged variable). Additionally I want to check on autoregressive effects (I left that out in the syntax below). The problem is in the mediations. MPlus has a problem to run over the random effects "pM11 pM12 pM21 pM22". There occur two problems: First there is an error if there is no Mplus-lagged variable on the right side of the ON statement (if the variable is lagged by MPlus it is no problem). Second MPlus does not recognize the labeling "b" and "d" if I put those relations on random.
I hope you can help with my problems.

Thanks in advance.
Btw on the between level all variables and random effects are correlated.

!Mediation 1 (Coping Mediator)
pM11 | HSCL ON Intra_L1 (a);
pM12 | Intra_L1 ON Inter_L1&1 (b);
pM13 | HSCL ON Inter_L1&1;

!Mediation 2 (Alliance Mediator)
pM21 | HSCL ON Inter_L1 (c);
pM22 | Inter_L1 ON Intra_L1&1 (d);
pM23 | HSCL ON Intra_L1&1;

Inter_L1 WITH Intra_L1;
p23 | Intra_L1 ON HSCL&1;
p13 | Inter_L1 ON HSCL&1;
 Bengt O. Muthen posted on Wednesday, November 15, 2017 - 10:34 am
You should not put a label like

pM11 | HSCL ON Intra_L1 (a);

because this expression does not refer to one parameter. The parameters related to this random slope appear on the Between level and there can be several of these parameters (mean, variance, regression slope).

You mention lagged variables - how many time points do you have?

You cannot have two different, contradictory models in the same Model command. E.g. "intra" cannot be both a mediator and an IV.
 Julian Rubel license number: SABC60080610 posted on Thursday, November 16, 2017 - 7:08 am
Hi Bengt,

Thanks for your quick response.

If I leave out pM11, pM12, pM21 and pM22, the model runs perfectly fine. If I put pM11 or pM21 into the model, I get this Error report:

*** ERROR in MODEL command Models with random slopes for dependent variables with missing values cannot be estimated with the BAYES estimator unless the variable is defined as WITHIN. One or more observations has missing values for a dependent variable on the right-hand side of an ON statement in a random slope definition.

Is there any possibility to put these relations on random? And what is the reason that this does not work? I quite do not understand this yet.

If I put pM12 or pM22 into the model, I get this report:

Unknown parameter label in MODEL CONSTRAINT: B

I wonder why this works when I fix the parameter instead of putting them on random. What do I have to do to change that?

I have three time points in total in the model, albeit there are only two time points in MPlus because one variable is manually lagged in SPSS before the MPlus analysis. The reason was that I had problems when I put a lagged variable before an ON statement. Therefore, each variable in this MPlus-model has only one lag. There are 1645 subjects with 4 to 113 observations.

I really appreciate your help.

 Bengt O. Muthen posted on Thursday, November 16, 2017 - 3:45 pm
Please send the 2 outputs to Support along with your license number.
Back to top
Add Your Message Here
Username: Posting Information:
This is a public posting area. Enter your username and password if you have an account. Otherwise, enter your full name as your username and leave the password blank. Your e-mail address is optional.
Options: Enable HTML code in message
Automatically activate URLs in message