Student and Subject Fixed effects

[Angie Heng]

Hi,

I am running a panel data with 3000 observations and each student has two subjects recorded.

I would like to run both student and subject fixed effects.

My code was

xtset STU_ID subject
xtreg testscore i.STU_ID i.subject SubjGradeA, fe cluster(STU_ID)

However, stata omitted all the dummy variables for STU_ID due to collinearity.
Does that mean that I do not require i.STU_ID for student fixed effect?

Thanks

[Carlo Lazzaro]

Angie:
as expected, Stata tells you that, being -xtset- with -STU_ID- as the panelid, -i.STU-ID- is collinear with the panel-wise fixed effect (please note that the -fe- machinery wipes out any time-invariant variable).
Hence, you should remove that predictor from the right-hand side of your regression equation.

[Angie Heng]

Does that imply that this code runs both student and subject fixed effect?

xtreg testscore i.subject SubjGradeA, fe cluster(STU_ID)

Whereas this code runs student fixed effect only?
xtreg testscore SubjGradeA, fe cluster(STU_ID)

[Carlo Lazzaro]

Angie:
the only fixed effect is the one you plug in -xtset- (that is, STU_ID).
Your other categorical variables are preditors that adjust each other in coefficients estimate.
If you want more than one -fe- (but I do not think that this is what you're after), you may want to take a look at the user-written command -reghdfe-.