Decomposing a covariate
Message/Author
 subrahmaniam tangirala posted on Friday, September 18, 2009 - 9:05 pm
I have an observed variable (i.e., X).

I wanted to decompose this variable into two latent variable parts (i.e., Xwij and Xbj) in my two-level model, and examine the effects of the level 2 assessment of X on my level 1 dependent variable (e.g., Ludkte et al., 2008, Psych Methods).

Hence, following the user manual (pg 230-231), I did not list X either in the within or the between statements. This allowed me to model X both in the within and between levels of my two-level model. Consequently, I was able to examine how level-2 assessment of x (i.e., Xbj) has a main effect on my level-1 Dependent variable.

However, now, I want to, go further, and test the interaction effect of the level-2 latent variable part (i.e., Xbj) with an observed between-level variable (e.g., Z) on my level-1 dependent variable.

Can you kindly guide me on how to test this interaction.

I tried creating a product term between X (where X is not specified as either a within or a between variable) and Z (specified as a between variable) using XWITH. But, I got an error message stating that I need to use the DEFINE function. However, I am not sure that I want to use the DEFINE function because I want to specifically create an interaction term between the level-2 "latent" assessment of X (i.e., Xbj) and an observed variable.

Any help would be much appreciated
 Linda K. Muthen posted on Saturday, September 19, 2009 - 8:46 am
Try the following where I put a latent variable behind the between part of x that is identical to the between part of x:

MODEL:
%WITHIN%
u ON x;
%BETWEEN%
f BY x; x@0;
w ;
int | f XWITH w;
u ON w f int;
 subrahmaniam tangirala posted on Saturday, September 19, 2009 - 6:08 pm
Thanks, Linda! This solution worked.
 Nicole Nugent posted on Sunday, May 16, 2010 - 9:02 am
I would like to decompose an observed interaction term that is predicting intercept in a growth model. In other software programs, I have done this ala Jaccard & Turkesi using variance, covariance, and unstandardized slopes (B) etc to plot the slopes of high and low levels of the components of the interaction term. Most of these values appear easily accessed in MPlus output (assuming that I can use the unstandardized Estimate term as B values). For some reason, I'm having trouble ascertaining the unstandardized B regression model constant. Can you please point me to where to find this output or to another way to decompose my interaction?

Thank you in advance!
 Linda K. Muthen posted on Monday, May 17, 2010 - 10:01 am
The unstandardized intercept is under Intercepts in the results. If you are using an older version of the program, you may need to add TYPE=MEANSTRUCTURE to the ANALYSIS command.
 Nicole Nugent posted on Wednesday, May 19, 2010 - 6:08 am
Thanks for your help - indeed it was clearly indicated when I loaded my Version 5 - along with many other helpful points if information (i.e., p values, additional error information, etc). Thanks again for this amazing software and for your generosity with your time and skill.
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: