on linux, the regression results using do file and run the command in a single line, get different results, wierd

[Fred Lee]

As shown, the log below shows the regression is normal with p values; however, when I run the syntax using command window, the regression resutls have no p values, shown as the screenshot, with the same coefficients. Why?
I tried on Mac, it also indicates warning: variance matrix is nonsymmetric or highly singular.
I am curious why run the syntax in a do file get normal results while others not?

> cap /* noisily */ reghdfe Evaluation c.divergent_assoc_fin_zici##c.lnwordcount $ctrl `tuple`i'' if AIorNot ==`k',vce(cluster firmID) absorb(ind )
> if _rc==0 {
> frame post zsignificance$num ("`k'") ("Evaluation") ("`tuple`i''") ("c.divergent_assoc_fin_zici##c.lnwordcount") (el(r(table),1,3)) (el(r(table),4,3)) //("active female")
> (el(r(table),1,11)) (el(r(table),4,11)) (`pdiff') /* (el(r(table),1,30)) (el(r(table),4,30)) */
> }
>
> } */
. }
16.
. }
(MWFE estimator converged in 1 iterations)

HDFE Linear regression                            Number of obs   =      4,905
Absorbing 1 HDFE group                            F(  18,    980) =       2.15
Statistics robust to heteroskedasticity           Prob > F        =     0.0036
                                                  R-squared       =     0.0110
                                                  Adj R-squared   =     0.0059
                                                  Within R-sq.    =     0.0055
Number of clusters (firmID)  =        981         Root MSE        =     9.5687

                               (Std. err. adjusted for 981 clusters in firmID)
------------------------------------------------------------------------------
             |               Robust
  Evaluation | Coefficient  std. err.      t    P>|t|     [95% conf. interval]
-------------+----------------------------------------------------------------
divergent_~i |   19.80279   9.455932     2.09   0.036     1.246588    38.35899
 lnwordcount |   157.3122   80.21144     1.96   0.050    -.0936979    314.7182
             |
          c. |
divergent_~i#|
          c. |
 lnwordcount |  -1.870194   .9563166    -1.96   0.051    -3.746857    .0064702
             |
     firmAge |   -.285815   .1006824    -2.84   0.005     -.483393    -.088237
   lnEmplNum |   .2929572   .2424283     1.21   0.227    -.1827811    .7686956
             |
  prodPhaseA |
Prototypi..  |  -.0004109    .436924    -0.00   0.999    -.8578252    .8570035
Manufactu..  |  -.1722763   .4932308    -0.35   0.727    -1.140186    .7956338
             |
   standardB |  -.1800625   .3052489    -0.59   0.555    -.7790792    .4189541
     award01 |   .7522663   .3552257     2.12   0.034     .0551758    1.449357
  lnTtlAsset |    .141961   .0938348     1.51   0.131    -.0421793    .3261012
emplRatioA~l |  -.8827885   .9497246    -0.93   0.353    -2.746516    .9809393
 emplRatioRD |    1.58624    .838761     1.89   0.059     -.059734    3.232214
         DAR |  -.0031675   .0042393    -0.75   0.455    -.0114866    .0051516
    devFormB |   .2795039   1.012329     0.28   0.783    -1.707078    2.266085
             |
   oprPhaseB |
Initial o..  |   .4197171   .5235594     0.80   0.423    -.6077093    1.447144
 Mass sales  |   .9096955   .7092001     1.28   0.200    -.4820301    2.301421
             |
 RDComplness |  -.1634139   .8766622    -0.19   0.852    -1.883765    1.556937
RDSelfInputR |   1.602693   1.562455     1.03   0.305     -1.46345    4.668836
       _cons |  -1594.692   793.2084    -2.01   0.045    -3151.274   -38.10931
------------------------------------------------------------------------------

[Fred Lee]

on another linux server, the same syntax on Mac has the p-values, while the same syntax has no p values on Linux. The version of reghdfe is the same and all Stata is the latest version

[Joseph Coveney]

As shown, the log below shows the regression is normal with p values; however, when I run the syntax using command window, the regression resutls have no p values, shown as the screenshot, with the same coefficients. Why?

Well, for one, you're not using the same syntax between the two: you're using capture in front of the command in the first case and you're not in the second case.

Is is possible that the capture prefix in the putatively successful first case is forcing the code to ignore an exception raised by a called program that checks for asymmetric or singular variance-covariance matrix and allowing things to proceed to computing the (erroneous) standard errors?

Regardless, I suspect with a little patience you can get the answer to the question you raised by means of set trace on. In the meantime you might want to take those standard errors with a grain of salt.

Second, it seems that the command that you're having trouble with is user-written and is not official Stata. You'll probably want to point that out in future posts.