Message/Author 

Hong Deng posted on Saturday, June 16, 2012  12:12 am



Dear Prof. Muthen, I tested a 111 multilevel mediation model with the syntax below. It worked well. Then I tried to add an interaction effect between another level1 moderator w and m on y by using the DEFINE function. mplus indicated that the interaction term i coined in DEFINE function is unknown. I also tried the ALGORITHM = INTEGRATION function, it didn't work either. My question is what is the right way to modify the syntax below if i want to include an level moderator. Thanks very much in advance. TITLE: 111 mediation (MSEM) DATA: FILE IS mydata.dat; VARIABLE: id x m y; CLUSTER IS id; ANALYSIS: TYPE IS TWOLEVEL RANDOM; MODEL: %WITHIN% m ON x(aw); y ON m(bw); y ON x; %BETWEEN% x m y; m ON x(ab); y ON m(bb); y ON x; NEW(indb indw); indw=aw*bw; indb=ab*bb; OUTPUT: TECH1 TECH8 CINTERVAL; 


If you want to use a variable created in DEFINE in the analysis, you need to put it at the end of the USEVARIABLES list after the original variables. 

Hong Deng posted on Saturday, June 16, 2012  8:30 pm



Dear Prof. Muthen, Thank you very much for the answer. I have a followup question. Do I need to model the interaction effect at BETWEEN level or I just simply put the level 2 variance of the interaction term there? 


The interaction terms should be created in DEFINE not MODEL CONSTRAINT. The within interaction should be used on within and the between interaction should be used on between. 

Eric Deemer posted on Wednesday, October 02, 2013  8:16 am



Hi, I'd like to do a likelihood ratio test on the difference between 2 nested multilevel mediation models at 1 df difference using MLR estimation. I'd like to do this test using the log likelihoods but it seems I need the scaling correction factor to calculate this. My question is, how do I request the scaling correction factor? Is this available using "type = twolevel random"? Eric 


You will get the scaling correction factor in the output with the loglikelihood. If it is not in the output, please send the output and your license number to support@statmodel.com. 

Back to top 