

Mediated Moderation with Count NBI Ou... 

Message/Author 


Hello, I’m currently working on a mediated moderation model with the following structure: Sentence on Responsibility(CntR1); Sentence#1 on Responsibility(InfR1); Responsibility on NN(R2); NN on Defect (b1) Prime (b2) PxD(b3); I used Model Constraint to calculate the simple slopes. lSS = (b1 + b3*(.5));!FW hSS = (b1 + b3*(.5));!SD Question 1. Is it reasonable to compute the indirect effect of the simple slopes on Sentence? !Sentence Count cLIND = (lSS*R2*CntR1);!FW cHIND = (hSS*R2*CntR1);!SD !Zero Inflation zLIND = (lSS*R2*InfR1);!FW zHIND = (hSS*R2*InfR1);!SD I’ve seen elsewhere on the discussion boards that indirect effects for these kinds of count variables can be calculated by using Model Constraint with S.E. and tests for that product of coefficients is all handled by Mplus. What I haven’t seen is this treatment of the simple slopes within Mplus as an indirect effect. That is can I exponentiate cLIND to zHIND and meaningfully interpret them. Thank you. 


Sorry, you lost me with the b3*(.5) expression. 


Sorry for being unclear. Defect (B1), Prime(B2), and PxD(B3) are two experimentally manipulated variables and their interaction: 2(Neurological Defect: Drug Induced, Congenital) x 2(Prime: Free Will, Scientific Determinism). .5 and .5 represent the effect coding that I used for the variables. The two simple slopes represent the effect of the Defect manipulation on NN for the Free Will (.5) and Scientific Determinism (.5) Prime manipulations respectively. 


In general in count models, an indirect effect is an effect on the log rate, which means that the indirect effect can be exponentiated into a rate. The same would hold for the inflation part I would assume. 

Back to top 

