Cross-level moderator PreviousNext
Mplus Discussion > Multilevel Data/Complex Sample >
Message/Author
 Virpi poyhonen posted on Friday, August 07, 2009 - 4:33 am
I want to test a cross-level interaction where I expect the relation of two individual level variables to be moderated by the association of two other variables. This association would be a group level variable (classroom). For instance, if there is a strong positive association between two variables in certain classrooms is the relation btw individual level variables weakened? I don’t know what would be the best way to do this. Can I use an aggregate of correlation as the classroom level moderator variable or is there a better way to do this?
 Linda K. Muthen posted on Friday, August 07, 2009 - 9:54 am
Cross-level interactions are specified as random slopes in Mplus. See Example 9.2 and slide 43 of the Topic 7 course handout for details.
 Ayesha Delany-Brumsey posted on Tuesday, May 11, 2010 - 2:35 pm
I want to test a cross-level interaction with one level 1 continuous variable and one level 2 continuous variable. I have complex survey data with weight, cluster, stratification, and sub-population. When I try to run the model I receive the following warning and error message:

*** WARNING in ANALYSIS command
The INTEGRATION option is not available with this analysis.
INTEGRATION will be ignored.
*** FATAL ERROR
THIS MODEL CAN BE DONE ONLY WITH MONTECARLO INTEGRATION.

I am having trouble figuring out if there is a way to run this model. Below is my model code. Thank you.:

Model:
%BETWEEN%
bpiint ON c_concdis c_SocCap;
s on c_concdis;
bpiint WITH s;
%WITHIN%
s| bpiint on c_NeighAttch;
bpiint on RSC_sex RSC_Psych;
RSC_sex RSC_Psych c_NeighAttch;
Analysis:
TYPE = COMPLEX TWOLEVEL RANDOM;
ESTIMATOR = MLR;
INTEGRATION = MONTECARLO;
 Ayesha Delany-Brumsey posted on Tuesday, May 11, 2010 - 5:57 pm
I am sorry, I realized I should have been more clear in my original post. I originally tried running the model with simply

Analysis:
TYPE = COMPLEX TWOLEVEL RANDOM;

But I got the same fatal error message.

Thank you,
Ayesha
 Linda K. Muthen posted on Wednesday, May 12, 2010 - 8:26 am
Try adding ALGORITHM=INTEGRATION; to the ANALYSIS command. If that does not help, please send the full output and your license number to support@statmodel.com.
 Ayesha Delany-Brumsey posted on Thursday, July 08, 2010 - 7:53 am
Hello I have a cross-level interaction with one level 1 continuous variable and one level 2 continuous variable. In the output, is the between level intercept of s the main effect of my level 1 variable? If not, where is the coefficient for the main effect for my level 1 variable?

Thank you,
Ayesha
 Linda K. Muthen posted on Friday, July 09, 2010 - 10:13 am
Yes.
 Mukadder  posted on Wednesday, June 01, 2011 - 6:53 am
Hi Dr Muthéns,
I'm trying out a two-level model including upper level mediation (2-2-1 mediation)among three latent variables. I wrote the syntax such that;
USEVARIABLES ARE f1 f2 f3;
WITHIN = f1;
BETWEEN = f2 f3;
CLUSTER IS class;
ANALYSIS:
TYPE = TWOLEVEL RANDOM;
MODEL:
%WITHIN%
f1 ON f2;
f1 ON f3;
%BETWEEN%
f2 ON f3;

Rightfully, the Mplus output gave an error because of the specification of the latents at the levels, specifically f2. Now I'm confused that the output gave me what I intended to do. Because I want to test the effect of f1 on the relationship between f2 and f3.
Your help is appreciated...Thanks
 Linda K. Muthen posted on Wednesday, June 01, 2011 - 10:40 am
I think what you want is the following. You would need to take f1 off of the WITHIN list.

%WITHIN%
f1 BY .....
%BETWEEN%
f1b BY ...
int | f1 XWITH f3;
f2 ON f3 int;
 Mukadder  posted on Wednesday, June 01, 2011 - 12:26 pm
Thank you very much Linda, so happy that this worked!
I have a follow up question...I tried a somehow multilevel path model. I created latent variables such that f1 is the sum of correct responses (German achievement), likely f2 (English achievement) and f3 (French achievement. f1 is a within variable; f2 and f3 are between variables. I followed the syntax you suggested to test the effect of f1 on the relationship between f2 and f3. However, Mplus told me that I can't use the XWITH command with the observed variables. In such a situation would the WITH command be more appropriate?
 Mukadder  posted on Wednesday, June 01, 2011 - 12:29 pm
Sorry I mistyped. The output told me that XWITH is appropriate for only observed variables not latents.

Thanks again.
 Mukadder  posted on Wednesday, June 01, 2011 - 12:44 pm
Linda, I think I have to learn how to read before how to use Mplus!
The output for my trial multilevel path model says:


The XWITH option is not available for observed variable interactions. Use
the DEFINE command to create an interaction variable.
Problem with: INT | f1 XWITH f3
 Linda K. Muthen posted on Wednesday, June 01, 2011 - 3:11 pm
Yes, so use the DEFINE command to create the interaction instead of the XWITH option.
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: