r(3021) error on ppmlhdfe

[George-Sun-ECON]

I am running a structural gravity model. However, after updating the package of ppmlhdfe and ftool, the following code does not report the regression table anymore, which was working on previous version:

ppmlhdfe trade1 x fta, a(i.exp_id#i.imp_id i.exp_id#i.year i.imp_id#i.year) vce(cluster i.exp_id#i.imp_id)  

The error info is as follows:

HDFE PPML regression                              No. of obs      =    391,691
Absorbing 3 HDFE groups                           Residual df     =     18,839
Statistics robust to heteroskedasticity           Wald chi2(2)    =      26.13
Deviance             =  1.04753e+10               Prob > chi2     =     0.0000
Log pseudolikelihood =  -5239046567               Pseudo R2       =     0.9939

Number of clusters (exp_id#imp_id)=    18,840

_b_table::init_table():  3021  class compiled at different times
_b_table::report_table():     -  function returned error
           _coef_table():     -  function returned error
                 <istmt>:     -  function returned error
r(3021);

I am wondering hwo to fix it. Thanks so much.

[sergiocorreia]

According to a quick google search, this feels like a stata issue. Maybe the following help:

• Ensure you have the latest versions of both ftools, reghdfe, and ppmlhdfe from github not SSC (use which to verify)
• Run ftools, compile to ensure the mlib file created by ftools is up-to-date
• Else, LMK which version of Stata you are using

[tomzylkin]

Hi, I wonder if it is the use of i. where you should just be using the names of the id variables.

I would code what you are trying to do as follows:
ppmlhdfe trade1 x fta, a(exp_id#imp_id exp_id#year imp_id#year) vce(cluster exp_id#imp_id)

Does that solve your problem?