I am running a mult. med. model (2 meds.) utilizing the half-long. tip that is described in Cole & Maxwell (2003).I'm having trouble figuring out the syntax to use(been trying to modify the syntax given in Preacher & Hayes (2008)). How would one specify the indirect effects? Also, how could one compare the size of these indirect effects across groups? Here is the syntax I've been using: ... GROUPING IS aacauc (0 = EA 1 = AA); ANALYSIS: BOOTSTRAP = 5000; MODEL: survt2 on survt1; shamet2 on survt1 shamet1; anxt2 on survt1 anxt1; bdt2 on survt1 shamet1 anxt1 bdt1;
survt1 with shamet1 anxt1 bdt1; shamet1 with anxt1 bdt1; anxt1 with bdt1;
shamet2 on survt1 (a1); anxt2 on survt1 (a2); bdt2 on shamet1 (b1); bdt2 on anxt1 (b2); MODEL INDIRECT: bdt2 IND shamet2 survt1; bdt2 IND anxt2 survt1; MODEL CONSTRAINT: NEW (a1b1 a2b2 con); a1b1 = a1*b1; a2b2 = a2*b2; con=a1b1-a2b2; OUTPUT: CINTERVAL (BCBOOTSTRAP) **note: t1 and t2 indicate timepoint I'm receiving error messages that no indirect or direct paths exist in the model for bdt2 IND shamet2 (also for anxt2). This makes sense to me, but I don't know how to specify the indirect effects, given the half-long. design. Thanks! Ellen Fitzsimmons, UNC-CH
MODEL INDIRECT cannot calculate indirect effects if the regression coefficients are not estimated as part of the model. There is no regression specified that relates bdt2 and shamet2 or anxt2 and no indirect path relates them.