Covariance in two-level interaction
Message/Author
 Lieke ten Brummelhuis posted on Wednesday, October 30, 2013 - 2:37 pm
Hi,

I am working with diary data (days nested in persons) and I estimate a two-level interaction model. I wondered if I should include the covariance between the product term and the components as well. Thus, that would be for instance:
y ON x m xm;
xm WITH x m;

Thank you for your response.

Lieke ten Brummelhuis
 Bengt O. Muthen posted on Wednesday, October 30, 2013 - 8:40 pm
x, m, xm are covariates so they are automatically correlated; don't mention WITH.
 Lieke ten Brummelhuis posted on Thursday, October 31, 2013 - 8:00 am
Thank you for your quick reply. I actually ran into another problem. Is there an easy way to calculate cross products from group-mean centered variables? I read some of the previous discussions and they suggested to first GMC the data, save it and then calculate the cross-product using a define command. Is there a new way in the latest Mplus version to do this at once?

Thank you!
Lieke
 Linda K. Muthen posted on Friday, November 01, 2013 - 7:09 am
There is not available in Version 7.1. It will be available in Version 7.2.
 S REN posted on Saturday, February 24, 2018 - 9:40 am
Hi, I wonder when to add the WITH functions to estimate the covariance.

My Level 1 hypothesis is that
Y1 Y2 ON M2 M1 X;
s1 | M1 ON X;
s2 | M2 ON M1;

For Level 2, I hypothesise W moderates s1 and s2

Should I request the covariance
s1 WITH M1 W;
s2 WITH M2 W;

OR should I add the following:
s1 WITH M1 W s2;
s2 WITH M2 W;

or should I do the following?
s1 WITH M1 W s2 Y1 Y2;
s2 WITH M2 W;

The above ways to covariate level 2 variables actually change the results. So I wonder which is the proper way to do (the covariance amongst them is not my hypothesis).
Thank you.
 Bengt O. Muthen posted on Saturday, February 24, 2018 - 1:56 pm
I would have the level 2 model

y1 y2 m1 m2 s1 s2 on w;

y1 y2 m1 m2 a1 s2 with y1 y2 m1 m2 a1 s2;