I'm fitting a simple OLS-model including an interaction term. To interpret the interaction I compute AMEs:
Code:
reg ITL OS LFR GD IW CWL LS noempl i.sex i.poor_srh cheftot c.CV##i.function margins, dydx(CV) at (function=(1 2 3)) marginsplot
Code:
Average marginal effects Number of obs = 220 Model VCE : OLS Expression : Linear prediction, predict() dy/dx w.r.t. : CV 1._at : function = 1 2._at : function = 2 3._at : function = 3 ------------------------------------------------------------------------------ | Delta-method | dy/dx Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- CV | _at | 1 | .3672521 .1679337 2.19 0.030 .0361438 .6983604 2 | .0134818 .1664265 0.08 0.936 -.3146549 .3416185 3 | .4481599 .2125448 2.11 0.036 .0290936 .8672261
My question is a basic and general one - I'm wondering whether I can conclude that "the effect of CV depends on function group"? What makes me doubt such a conclusion is the fact that, while the effect of CV is only significant in two of the function groups, all three CIs overlap each other?
Best,
Caroline
0 Response to Interpretation of AMEs: CIs overlapping each other but not zero, can I conclude that there is an interaction?
Post a Comment