Multi-Level SEM with Latent Variables PreviousNext
Mplus Discussion > Multilevel Data/Complex Sample >
Message/Author
 Susanna Ho posted on Sunday, April 04, 2010 - 5:43 pm
Dear Drs Muthen,

I am working on a multi-level SEM. Both independent and dependent variables are latent construct.

In the model,

Latent_A (individual) and Latent_B (group) will affect Latent_C (individual)

I include the BY statement for Latent_C at the WITHIN level, and at the BETWEEN level, the variable cannot be seen. See the code below.

MODEL: %WITHIN%
LA BY LA1 LA2 LA3;
LC BY LC1 LC2 LC3;

LC ON LA;

%BETWEEN%
LB BY LB1 LB2 LB3;

LC ON LB; <====

There is an error message saying that "LC is unknown". Probably because I create it at the WITHIN level.

Please advise what I should do.

Thanks a lot.
 Linda K. Muthen posted on Monday, April 05, 2010 - 8:36 am
You need to define LC on both levels using different names for the factor, for example, LCW and LCB. The factor loadings can be held equal across levels if this fits the data.
 Sabrina Krys posted on Monday, January 27, 2020 - 8:23 am
Hi!

I am conducting a multilevel mediation SEM with random intercepts and fixed slopes. However, a reviewer of my current paper suggested to test for random slopes and to groupmean center the predictor variable. Unfortunately, I think it is not possible to test for random slopes without montecarlo integration (and I really don't know how to use this) and the models I tried previously couldn't be computed due to low memory capacity. I also don't know how to groupmean center a latent variable in Mplus.

This is an extract of my current model in Mplus 8.4:

ANALYSIS:
type=twolevel;
estimator=ML;

Model:
%within%
f1w by item1 item2 item3 item4;
f2w BY item1 item2 item3;
f3w BY item1 item2 item3;

f2w ON f1w;
f3w ON f1w f2w;

%between%
f1b by item1 item2 item3 item4;
f2b BY item1 item2 item3;
f3b BY item1 item2 item3;

f2b ON f1b;
f3b ON f1b f2b;

MODEL INDIRECT:
f3w IND f1w;


My questions are:
1. How can I test if the slopes of the particular effects are random?
2. How can I groupmean center the predictor variable f1w?

Kind regards
Sabrina Krys
 Bengt O. Muthen posted on Monday, January 27, 2020 - 4:42 pm
The memory requirement is a function of the number of dimensions of integration and the sample size. To guide you, you need to send your data output with random slopes - along with your license number.
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: