3-2-1 Mediation Model with binary out...
Message/Author
 Claire Ryu posted on Thursday, May 31, 2018 - 9:01 am
I am new in Mplus.
I need to perform a mediation analysis for 3-2-1.
Can Mplus perform a 3-level SEM with binary outcome to test 3-2-1 mediation? Or is something else is available to test 3-2-1 mediation?

X: continous (level 3)
M: mediator, continous (level 2)
Y: outcome, binary (level 1)

I also want to know how to include predictors at each level.
Thank you so much!!!
 Tihomir Asparouhov posted on Thursday, May 31, 2018 - 11:23 am
Here is some sample code for your situation

variable:
names are y m x c2 c3;
between= m;
between=(c3) x;
categorical=y;
cluster=c3 c2;

data: file=1.dat;

analysis: type = threelevel;

model:
%between c2%
y on m; m;
%between c3%
y on m x;
m on x;
 Zhi Ye posted on Thursday, September 20, 2018 - 11:20 pm
Dear Dr Muthen,
I am running an 3-2-1 multilevel mediation model as following:
MODEL:
%WITHIN%
PVW by PV1@1 PV2 PV3 PV4;
PVW on gender ses age;
PVW on skill;
Skill;
%BETWEEN classid%
PVB1 by PV1 PV2 PV3 PV4;
PVB1@1; [PVB1@0];
PVB1 on prosocial;
PVB1 on skill;
skill on prosocial;
%BETWEEN schoolid%
PVB2 by PV1 PV2 PV3 PV4;
PVB2@1; [PVB2@0];
PVB2 on climate prosocial skill;
skill on prosocial climate;
prosocial on climate;
OUTPUT: TECH1 TECH8 STDYX;

However, I have got an error message:
Unrestricted x-variables for analysis with TYPE=THREELEVEL must be a variable on
only one level. The following variable cannot exist on more than one level:
PROSOCIAL

Could you help me to fix this problem, please?

Thanks so much!

Best
 Bengt O. Muthen posted on Friday, September 21, 2018 - 5:39 pm
A latent variable decomposition is not done with 3-level; you have to create the cluster variable yourself. If this doesn't help, send your full output to Support along with your license number.