Referencing first non-omitted level of factor variable?

[Ben Marrow]

I'm trying to put a title over a factor variable by using the refcat command of esttab. However, the first level of my variable is being ommitted due to collinearity in the regression, so when I try to use refcat(1.category "Category Title") , it does not know what 1.category is and omits the title altogether.

My code is:

reghdfe y ib(last).pair_c ib(last).pair_c#grade , a(week)
esttab using "out.tex", refcat(1.pair_c "Pair Type")

Manually, in the above code, I can change 1.pair_c to 2.pair_c and it will work. However, is there a way I can dynamically have stata figure out the first non omitted level of pair_c?

[Clyde Schechter]

Try this approach:

sysuse auto, clear

//    CREATE A COLINEARITY INVOLVEING 1.rep78
gen junk = 1.rep78

regress price mpg i.junk ib(last).rep78

matrix B = e(b)
local bnames: colnames B
display `"`bnames'"'

//    FIND FIRST NON-OMITTED LEVEL OF rep78
levelsof rep78, local(levels)
foreach l of local levels {
    if `:list posof "`l'.rep78" in bnames' {
        local foundit `l'
        continue, break
    }
}
display `foundit'

The logic is that omitted levels do not appear simply as #.varname in the column names of e(b), they appear as #o.varname or #b.varname. So loop over the levels of the variable of interest until you find one for which #.varname does appear in e(b)'s column names..