ppmlhdfe error when no independent variable is included

[Harry Li]

Dear STATA community,

I run into error when I run ppmlhdfe with no independent variable. For example:

use "http://fmwww.bc.edu/RePEc/bocode/e/EXAMPLE_TRADE_FTA_DATA" if category=="TOTAL", clear
ppmlhdfe trade , absorb( isoexp isoimp )

The error is:

                  mean():  3301  subscript invalid
            GLM::solve():     -  function returned error
                 <istmt>:     -  function returned error
r(3301);

Does anyone else have the same problem? Any advice is much appreciated!

[Justin Niakamal]

ppmlhdfe is a community-contributed command and requires the reghdfe and ftools packages, so make sure you have the latest versions installed. I am not able to reproduce your error using your code.

. 
. use "http://fmwww.bc.edu/RePEc/bocode/e/EXAMPLE_TRADE_FTA_DATA" if category=="TOTAL", clear
(Example gravity data for ppml_panel_sg (35 countries, 1988-2004, every 4 yrs))

. 
. ppmlhdfe trade , absorb( isoexp isoimp )
Iteration 1:   deviance = 2.4136e+10  eps = .         iters = 5    tol = 1.0e-04  min(eta) =  -3.80  PS  
Iteration 2:   deviance = 1.9549e+10  eps = 2.35e-01  iters = 3    tol = 1.0e-04  min(eta) =  -4.98   S  
Iteration 3:   deviance = 1.9248e+10  eps = 1.56e-02  iters = 3    tol = 1.0e-04  min(eta) =  -5.89   S  
Iteration 4:   deviance = 1.9242e+10  eps = 3.27e-04  iters = 3    tol = 1.0e-04  min(eta) =  -6.24   S  
Iteration 5:   deviance = 1.9242e+10  eps = 2.04e-06  iters = 2    tol = 1.0e-04  min(eta) =  -6.28   S  
Iteration 6:   deviance = 1.9242e+10  eps = 3.25e-10  iters = 2    tol = 1.0e-05  min(eta) =  -6.28   S  
Iteration 7:   deviance = 1.9242e+10  eps = 1.33e-16  iters = 1    tol = 1.0e-07  min(eta) =  -6.28   S  
Iteration 8:   deviance = 1.9242e+10  eps = 0.00e+00  iters = 1    tol = 1.0e-09  min(eta) =  -6.28   S O
------------------------------------------------------------------------------------------------------------
(legend: p: exact partial-out   s: exact solver   h: step-halving   o: epsilon below tolerance)
Converged in 8 iterations and 20 HDFE sub-iterations (tol = 1.0e-08)

HDFE PPML regression                              No. of obs      =      5,950
Absorbing 2 HDFE groups                           Residual df     =      5,881
                                                  Wald chi2(0)    =          .
Deviance             =  1.92421e+10               Prob > chi2     =          .
Log pseudolikelihood =  -9621117926               Pseudo R2       =     0.6861
------------------------------------------------------------------------------
             |               Robust
       trade |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
       _cons |   16.07992   .0337047   477.08   0.000     16.01386    16.14598
------------------------------------------------------------------------------

Absorbed degrees of freedom:
-----------------------------------------------------+
 Absorbed FE | Categories  - Redundant  = Num. Coefs |
-------------+---------------------------------------|
      isoexp |        35           0          35     |
      isoimp |        35           1          34     |
-----------------------------------------------------+

. 
end of do-file

. which ppmlhdfe 
/Users/justinniakamal/Library/Application Support/Stata/ado/plus/p/ppmlhdfe.ado
*! version 2.2.0 02aug2019
*! Authors: Sergio Correia, Paulo Guimarães, Thomas Zylkin
*! URL: https://github.com/sergiocorreia/ppmlhdfe

[Harry Li]

Thanks for letting me know! This is confusing--I upgraded ftools and reghdfe to the latest version, though the problem does not go away. Any idea what procedure I should follow to identify its cause? Thanks!

[Joao Santos Silva]

I also have no problem with this. Did you also update Stata and ppmlhdfe?

[Harry Li]

ppmlhdfe is the latest version. I'm using Stata/MP 14.0.

[Joao Santos Silva]

Even using version control, I cannot replicate your problem. maybe Sergio Correia can help.

[Sergio Correia]

I would that since ppmlhdfe depends on reghdfe and ftools, it might help if you update these two to their latest versions from Github (also, latest version of ppmlhdfe on github is 2.3.0 rather than 2.2.0)