Moderation with Cross-Lagged Panel Models

[Lourah Kelly]

I ran a cross-lagged panel model, testing bi-directional and longitudinal relationships among three variables (i.e., drinking frequency, stressful life events, and depressive symptoms) across four years of annual data in college students. There's a good deal of literature suggesting relationships among drinking and stress, and among drinking and depression are different for men and women. I'm having a hard time finding how to test for moderation with a binary variable (gender was measured as either male or female in this dataset) with a cross-lagged panel model with three variables over time. I'm thinking there might be moderation, since I split the file and tested the same CLPMs among only men and only women - and there are some paths that are only statistically significant among men and some only among women - but I don't actually know if gender is really a moderator (what I did was more of a quick way I knew how to test differences between men and women). Any examples of moderation with a binary variable with CLPMs with *three* variables is much appreciated!

[Lourah Kelly]

. sem (W1LESS -> w2LESS, ) (W1LESS -> w2BDI, ) (W1LESS -> w2drkhav, ) (W1BDI -> w2LESS, ) (W1BDI -> w2BDI, ) (W1BDI -> w2drkhav, > 2LESS, ) (w1drkhav -> w2BDI, ) (w1drkhav -> w2drkhav, ) (w2LESS -> W3LESS, ) (w2LESS -> W3BDI, ) (w2LESS -> w3drkhav, > BDI -> W3BDI, ) (w2BDI -> w3drkhav, ) (w2drkhav -> W3LESS, ) (w2drkhav -> W3BDI, ) (w2drkhav -> w3drkhav, ) (W3LESS -> > 4BDI, ) (W3LESS -> W4LESS, ) (W3BDI -> w4drkhav, ) (W3BDI -> W4BDI, ) (W3BDI -> W4LESS, ) (w3drkhav -> w4drkhav, ) (w3drkhav > hav -> W4LESS, ), method(mlmv) vce(robust) standardized cov( W1LESS*W1BDI W1LESS*w1drkhav w1drkhav*W1BDI e.w2LESS*e.w2BDI > .w2BDI*e.w2drkhav e.W3LESS*e.W3BDI e.W3LESS*e.w3drkhav e.W3BDI*e.w3drkhav e.W4BDI*e.w4drkhav e.W4LESS*e.w4drkhav e.W4LESS*note: Missing values found in observed exogenous variables. Using the noxconditional behavior. Specify the forcexcondition for this this behavior.
Endogenous variables
Observed: w2LESS w2BDI w2drkhav W3LESS W3BDI w3drkhav w4drkhav W4BDI W4LESS
Exogenous variables
Observed: W1LESS W1BDI w1drkhav

Could I potentially estimate moderator with a line from gender to the paths estimated from Time I to Time 2 and from Time 2 to Time 3 and from Time 3 to Time 4?