CSDID time fixed effects

[Nay Iturbe]

I understand that because of its design, time-fixed effects are already incorporated in the CSDID for staggered setups. However, if I add time dummies as independent variables in my model, the parallel trends are more well-defined. Does anyone know why is that? Can I include the time dummies as independent variables or this will bias my model?

[FernandoRios]

The most likely case, its a fluke caused by errors in the modeling.
In fact, if i add time fixed effects on the example i have in the help file, nothing is estimated at all.
Do not add time dummies.