Jing Liu posted on Wednesday, February 01, 2017 - 1:47 pm
Dear Drs Muthen,
How are you?
I would like to ask you a question about a multilevel first-stage moderated mediation model within a MSEM framework. I am testing whether the effect of employee orientation (CORNT, assessed at level 1) on employee satisfaction (ESat, level 2) via employee incivility (EINC, level 1) varies across levels of organizational formalization (ORGFORM, level 2). It is a 2(2-1-2) MSEM model with random slopes.
I ran the mediation model first, and then the moderation, and finally combined the two into a moderated mediation model. It turned out that both the mediation and moderated models ran successfully. However, there was some problems with the moderated mediation model. Here is the warning message: THE ESTIMATED WITHIN COVARIANCE MATRIX COULD NOT BE INVERTED. COMPUTATION COULD NOT BE COMPLETED IN ITERATION 195. CHANGE YOUR MODEL AND/OR STARTING VALUES.
Jing Liu posted on Wednesday, February 01, 2017 - 1:49 pm
The following is the syntax: ANALYSIS: TYPE IS TWOLEVEL RANDOM; Estimator = MLR; INTEGRATION = MONTECARLO(1000); !missing data on mediators ALGORITHM IS INTEGRATION; !moderation
MODEL: %WITHIN% CSATw BY CSAT1 CSAT2 CSAT3 CSAT4; EINCw BY EInc1 EInc2 EInc3 EInc4 EInc5 EInc6 EInc7; EINCw CSATw; !estimate the within-level variance sb | CSATw ON EINCw;
%BETWEEN% [sb] (wb1); EINCb BY EInc1 EInc2 EInc3 EInc4 EInc5 EInc6 EInc7; EORNTb BY EOrnt1 EOrnt2 EOrnt3 EOrnt4 EOrnt5; ORGFORMb BY OrgForm1 OrgForm2 OrgForm3 OrgForm4 OrgForm5; CSATb BY CSAT1 CSAT2 CSAT3 CSAT4; EORXORG | EORNTb XWITH ORGFORMb; EORNTb WITH ORGFORMb; EINCb ON EORNTb (ba1); EINCb ON ORGFORMb (ba2); EINCb ON EORXORG (ba3); CSATb ON EINCb (bb1); CSATb ON EORNTb (bb2); EINCb CSATb ORGFORMb EORNTb sb;
MODEL CONSTRAINT: NEW(b); b=wb1 + bb1;
OUTPUT: TECH1 TECH8 CINTERVAL SAMPSTAT;
I would greatly appreciate if you provide suggestions on how to make the coding work!