Predict residuals after running a FE / RE model

[Sang-Bum Park]

How can residual centering be done after running a panel data regression?

It can be done easily after running a simple linear regression as follows.

sysuse auto, clear
generate interact = weight*length
quietly regress interact weight length
predict interact_rc, res

However, after running a fixed effects model, the option residuals (res) does not work.
What option should be included at the end of the predict command instead of res for panel data regressions?

[Carlo Lazzaro]

Sang-Bum:

predict epsilon, e

Will probably give you back what you're intetested in.
See for more details -xtreg postestimation- entry in Stata .pdf manual.

[Sang-Bum Park]

Thank you, Carlo Lazzaro for your answer.

Is the below code correct for residual centering of the abovementioned interaction term?

predict interact_rc, e

[Carlo Lazzaro]

Sang-Bum:
Yes.

[Sang-Bum Park]

Thank you so much, Carlo Lazzaro.