t.-test for coefficients of the same regression

[Alberto Alvarez]

Dear Stata Users,

I am using the following regression:

reghdfe y x1 x2 x3 if inrange(fyear,2003, 2014), absorb(gvkey_destr fyear) vce(cluster gvkey_destr fyear)

what I want is to check if x2+x3 is significantly different from zero. I used the following code:

test [x2 + x3]   = 0

Is that right? The problem is that with a pooled ols I was using the following code, but “reghdfe” does not support it:

xi: reg y x1 x2 x3 if inrange(fyear,2003, 2014)
estimate store c_zero_opport
suest c_zero_opport, vce(cluster gvkey)
test [mean]x2   +  [mean]x3   = 0

[Clyde Schechter]

test [x2 + x3] = 0

will work. You don't actually need the square brackets there, but they don't hurt anything either.

Note: in your original

xi: reg y x1 x2 x3 if inrange(fyear,2003, 2014)
estimate store c_zero_opport
suest c_zero_opport, vce(cluster gvkey)
test [mean]x2 + [mean]x3 = 0

you have made things much more complicated than is necessary. First, the -xi:- prefix does nothing there. (You probably shouldn't be using it anyway, as for almost all purposes factor-variable notation is better.) Next, there is no need to go through storing the estimates and running them through -suext-. All you needed is:

reg y x1 x2 x3 if inrange(fyear, 2003, 2014)
test x2 + x3 = 0

to do that.

General point: -suest- is only needed to test coefficients across different regressions. For testing things about coefficients in a single equation, just go directly to -test-.

[Alberto Alvarez]

Thank you very much!