Stata encountering negative weights, but no negative weights in data!!

[Dana Shaat]

Hi everyone,

I am trying to run a weighted regression using reghdfe. Here is my code:

local filename "an1"
local X0 "oberegion#year year unitid" //fixed effects
local eventtime exp1
local sample "if ever_certonly!=1 & ever_not24yr!=1 & ever_notpub!=1 & ever_special_use!=1"


preserve
set more off
local r replace
forvalues s=0/1{    
        if `s'==0 {
            local spec " & iclevel ==1 | treated == 1"
        }
        else if `s'==1 {
            local spec " & iclevel==2 | treated == 1 "
        }
        reghdfe log_eftotlt_ftf _T_`eventtime'* `sample' `spec' [w=wt], a(`X0') cl(fips)
}

restore

And I am getting this error!

(analytic weights assumed)
weight wt can only contain strictly positive reals, but 25242 negative values were found!
r(402);

I double checking my weight variable (wt) and there are no negative values. Not sure why this error is coming up. Any ideas?

[daniel klein]

Thanks for providing code.

I double checking my weight variable (wt) and there are no negative values.

How exactly did you do that? Type

summarize wt
count
assert (wt > 0) & !mi(wt)

and show us the output.

[Bjarte Aagnes]

The community-contributed package reghdfe (ssc describe reghdfe) , which requires the community-contributed ftools package, is possibly returning an incorrect error message. See -help trace- to debug.