I'm working with a 2-1-1 mediation model (each number refers to the level of analysis) with a level 1 moderator of the 'b' path (i.e., between mediator and dependent variable). Mplus code is provided below.

The following link to a diagram (from Preacher, Rucker, & Hayes 2007) demonstrates what I'm after, but in a single-level context. In my case, X is at level 2, and all other variables are at level 1.

https://i.stack.imgur.com/sCvUi.pngc

I would appreciate input on the following:

1) Based on the given diagram, whether I should add certain WITH statements (corresponding to the curved arrows).

2) Whether the estimation of Level 1 and Level 2 (residual) variances is proper as specified.

3) Whether the decision to not group mean center variables (working from Lüdtke et al. 2008) is correct/advisable. (I still used group means for the Level 2 x variable; see DEFINE statement.)

Mplus code:

USEVARIABLES = GROUP X M W Y XGRPMN MW;

Missing are . ;

BETWEEN IS XGRPMN;

CLUSTER IS GROUP;

DEFINE:

XGRPMN = CLUSTER_MEAN (X);

MW = M*W;

ANALYSIS: TYPE IS TWO-LEVEL RANDOM;

MODEL:

%WITHIN%

M W MW Y; ! estimate L1 (residual) variances for m, w, m*w, y

Y ON M(b1); ! regress L1 y on L1 med, slope "b1"

Y ON W(b2); ! regress L1 y on L1 path 'b' mod, slope "b2"

Y ON MW(b3); ! regress L1 y on (L1 med*L1 path 'b' mod), slope "b3"

%BETWEEN%

XGRPMN M W MW Y; ! estimate L2 (residual) variances for x, m, w, m*w, y

M ON XGRPMN(a1); ! regress L1 med on L2 x, slope "a"

Y ON M; ! regress L1 y on L1 med

Y ON XGRPMN(cdash); ! regress L1 y on L2 x, slope "cdash"

MODEL CONSTRAINT:

NEW(LOW_MOD MED_MOD HIG_MOD

I_LOWMOD I_MEDMOD I_HIGMOD

IMM); ! Name indirect effects and Index of MM

LOW_MOD = 3.8755; ! example -1 SD of moderator

MED_MOD = 4.8596; ! example mean of moderator

HIG_MOD = 5.8437; ! example +1 SD of moderator

! Calc conditional indirect effects and IMM

I_LOWMOD = a1*(b1+b3*LOW_MOD); ! indirect effect cond on low value of mod

I_MEDMOD = a1*(b1+b3*MED_MOD); ! indirect effect cond on medium value of mod

I_HIGMOD = a1*(b1+b3*HIG_MOD); ! indirect effect cond on high value of mod

IMM = a1*b3; ! Index of moderated mediation