Multilevel mediation 2-1-1, Zero-infl...
Message/Author
 Jennifer Yahner posted on Wednesday, September 29, 2010 - 8:15 pm
Hi Drs Muthen & Muthen,

I am trying to estimate a multilevel mediation model 2-1-1 using code similar to Preacher's example E at website: http://www.people.ku.edu/~preacher/pubs/syntax_appendix_081009.pdf.

When I treat my dependent variable (level 1) as continuous, things work just fine.

But when I treat it as zero-inflated Poisson, or as Tobit, then I receive the error below--which basically disallows me from estimating the Between-group component of the 1-1 part of my 2-1-1 mediation path.

I would appreciate any advice/explanation - thank you!

-Jennifer

*** ERROR in MODEL command
Observed variable on the right-hand side of a between-level ON statement must be a BETWEEN variable.
 Linda K. Muthen posted on Thursday, September 30, 2010 - 12:12 pm
In the between part of the model command, if the variable is u, add

f BY u;
u@0;

and use f in place of u in the ON statement. This is simply a trick that makes Mplus think it is a between level variable. It has no statistical impact.
 Jennifer Yahner posted on Thursday, September 30, 2010 - 1:05 pm
Thank you so much! This code works, but now - of course - my model will not converge despite increased iterations.

I appreciate your help. I may need to email the support@statmodel with my data, but will try to correct things myself tomorrow morning.

Thanks!

-Jennifer
 Ok-young, JI posted on Monday, October 24, 2016 - 12:41 am
hi Dr. Mutehns
i have a model with multiple mediator
my model is consist of one level 2 variable (x1) and 2 level 1 mediator (m1, m4) and one level 1 dependent variable(y1)
my question is did i wrote a right syntax???
i can't find a 2-1-1-1 multilevel mediation model example..
my syntax is here.
data: file is norm_motive.txt ;
variable: names = x1 x2 m1 m2 m3 m4 y1 y2 w personalid team ;
usevariables = x1 m1 m4 y1 team ;
within = ;
between = x1 ;
cluster = team ;
analysis: type = twolevel random ;
estimator = ML ;
model: %WITHIN%
s1| m4 ON m1 ; ! within level p_DM_S -> WRDA
s2| y1 ON m4 ; ! within level WRDA -> p_PD
%BETWEEN%
m1 ON x1 (a) ; ! between level t_DSN -> p_DM_S
m4 ON m1 (s3) ; ! between level p_DM_S -> WRDA
y1 ON m4 (s4) ; ! between level WRDA -> p_PD
s1 WITH x1 m1 m4 y1 ;
s2 WITH x1 m1 m4 y1 ;

[s1](ms1) ; [s2](ms2) ;
model constraint:
new(b1 b2 ind) ;
b1=s3*ms1 ; ! copute between b1 path
b2=s4*ms2 ; ! compute between b2 path
ind=a*b1*b2 ; ! indirect effect t_DSN -> p_DM_S -> WRDA -> p_PD

output: sampstat tech3 tech8
 Bengt O. Muthen posted on Monday, October 24, 2016 - 2:38 pm
You are interested in mediation on Between. The means of the random slopes are only involved for Within-level indirect effects. The indirect effect on Between is simply a*s3*s4.