Moderation Analysis, omitted variables due to collinearity

[Stini Berg]

Hello,
I am trying to do a moderation analysis for a regression analysis which worked out perfectly well. However, doing the moderation and adding the moderator returns the messages:

note: 3bn.psa is probably collinear with the fixed effects (all partialled-out values are close to zero; tol = 1.0e-09)
note: 3.psa omitted because of collinearity

and the result table includes '0' for omitted.

The original regression command was:
reghdfe fundflows dummy 5 dummy1, vce(robust) absorb(Individual fund)

The moderation command is:
reghdfe fundflows i.dummy5##i.psa, vce(robust) absorb(Individual fund)

I tried to do the moderation command without FE (i.e. with noabsorb) which worked out, however, the outcome values changed their signs.


What is the problem with my moderation?

[Andrew Musau]

The variable "3.psa" is collinear with the fixed effects. So the coefficient on the variable cannot be identified in this model. If the variable is simply a control variable, just ignore the collinearity and proceed with your analysis (in other words, the effect of the variable is already captured by the fixed effects). If it is an independent variable of interest, then you have a problem with the model as is. reghdfe is from SSC (FAQ Advice #12).