Multilevel Mod-Med - L2 Interaction
Message/Author
 Ben Walsh posted on Monday, January 26, 2015 - 7:34 pm
I'm trying to test a 1-2-2 model, with a between-level dichotomous moderator (w) of the x-m relationship (all other var's continuous). My syntax (below) is based on Preacher et al. (10) and Preacher et al. (07).

A colleague suggested I try xwith since x is a latent variable at L2 and w is observed. However, I receive the errors below.

VARIABLE: NAMES ARE
group
w
other
m
y
other2
x;
USEVARIABLES ARE
group
w
m
y
x;
CLUSTER IS group;
BETWEEN ARE m y w;
ANALYSIS: TYPE IS TWOLEVEL RANDOM;
estimator = mlr;
ALGORITHM=INTEGRATION;
MODEL:
%WITHIN%
x;
%BETWEEN%
m y w;
xw | x xwith w;
y on m (b1)
x
w
xw;
m on x (a1)
w
xw (a3);

*** ERROR in MODEL command
The XWITH option is not available for observed variable interactions. Use
the DEFINE command to create an interaction variable.
Problem with: XW | X XWITH W

I've tried DEFINE command, but haven't had success because it identifies within variability in x (which there is, but I only want to specify the interaction at L2, using the x corrected for sampling error).
 Linda K. Muthen posted on Tuesday, January 27, 2015 - 1:54 pm
You can put a factor behind x on between and use the factor in the XWITH statement. Putting a factor behind x makes it equivalent to x.

f BY x@1; x@0;
 Ben Walsh posted on Tuesday, January 27, 2015 - 7:18 pm
Thank you for the feedback. I will give this a try.
 Ben Walsh posted on Wednesday, January 28, 2015 - 8:04 am
I used your suggestion with syntax below:

VARIABLE: NAMES ARE
group
w
other
m
y
other2
x;
USEVARIABLES ARE
group
w
m
y
x;
CLUSTER IS group;
BETWEEN ARE m y w;
ANALYSIS: TYPE IS TWOLEVEL RANDOM;
estimator = mlr;
ALGORITHM=INTEGRATION;
MODEL:
%WITHIN%
x;
%BETWEEN%
m y w;
fx BY x@1; x@0;
fxw | fx xwith w;
y on m (b1)
fx
w
fxw;
m on fx (a1)
w
fxw (a3);

However, now I receive: THE STANDARD ERRORS OF THE MODEL PARAMETER ESTIMATES MAY NOT BE TRUSTWORTHY FOR SOME PARAMETERS DUE TO A NON-POSITIVE DEFINITE FIRST-ORDER DERIVATIVE PRODUCT MATRIX. THIS MAY BE DUE TO THE STARTING VALUES BUT MAY ALSO BE AN INDICATION OF MODEL NONIDENTIFICATION. THE CONDITION NUMBER IS -0.133D-17. PROBLEM INVOLVING PARAMETER 14. Thank you for any feedback.
 Linda K. Muthen posted on Wednesday, January 28, 2015 - 8:14 am