Fixed Effects different from variables of xtset

[Manuel Matte]

Hi! I have a panel data quarterly (yq) and by firm (newid), so when i run xtset, y run

xtset newid yq

Now, i want to run the regression with xtreg but the fixed effect has to be by industry, with the variable sectorid and by yq. When i run

xtreg depvar indepvar i.yq i.sectorid, robust

automatically stata used randomm effects, and i want fixed effects. How can i change this?
If i try with ,fe Stata run fixed effetcs of newid.

[Andrew Musau]

xtset sectorid
xtreg depvar indepvar i.yq, fe robust

[River Huang]

Dear Manuel, You might to try (ssc install reghdfe and ssc install ftools):

reghdfe depvar indepvar, a(yq sectorid) vce(cl sectorid)

See

webuse grunfeld, clear

gen ind = 1
replace ind = 2 if company > 3
replace ind = 3 if company > 7

// Mine
reghdfe invest mvalue kstock, a(ind year) vce(cl ind)

// Andrew Musau
xtset ind
xtreg invest mvalue kstock i.year, fe robust

[Manuel Matte]

Thanks to both! And for the command reghdfe if i want to add robustness too, how can i do that? i tried with

vce(cluster sectorid, robust)

and

vce(robust) vce(cluster sectorid)

and didn't work, there's any way to do that in this command?

[Andrew Musau]

vce(robust) vce(cluster sectorid)

You cannot specify both at the same time. In reghdfe, -vce(robust)- gives you Huber-White standard errors and -vce(cluster clustervar)- gives you clustered standard errors. For panel data with \(N>>T\), the latter are robust to heteroscedasticity and arbitrary forms of serial correlation whereas the former are not appropriate. See the manual entry of xtreg for more discussion on this and the related literature.

help xtreg