r(2000) no observations when running xtreg

[Taz Raihan]

I have an unbalanced panel dataset. When I do the following regression, Stata produces output without any error:

 xtreg rgdpgrowth sxp sxp2 extdebtGDP i.year, fe vce(robust)

Fixed-effects (within) regression               Number of obs      =       315
Group variable: country                         Number of groups   =        65

R-sq:  within  = 0.1066                         Obs per group: min =         3
       between = 0.1082                                        avg =       4.8
       overall = 0.1036                                        max =         5

                                                F(7,64)            =      5.88
corr(u_i, Xb)  = 0.0506                         Prob > F           =    0.0000

                               (Std. Err. adjusted for 65 clusters in country)
------------------------------------------------------------------------------
             |               Robust
  rgdpgrowth |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         sxp |   -.043291    .097015    -0.45   0.657    -.2371007    .1505187
        sxp2 |   .0285514   .1760005     0.16   0.872      -.32305    .3801529
  extdebtGDP |  -.0000597   .0000177    -3.38   0.001     -.000095   -.0000245
             |
        year |
       1980  |  -.0117597   .0054863    -2.14   0.036    -.0227198   -.0007996
       1985  |  -.0206939   .0057291    -3.61   0.001    -.0321392   -.0092487
       1990  |  -.0123391    .005445    -2.27   0.027    -.0232167   -.0014615
       1995  |    -.01004   .0051349    -1.96   0.055    -.0202981    .0002182
             |
       _cons |   .0297318    .010096     2.94   0.004     .0095628    .0499008
-------------+----------------------------------------------------------------
     sigma_u |   .0212606
     sigma_e |  .02712922
         rho |  .38047986   (fraction of variance due to u_i)
------------------------------------------------------------------------------

But when I include a lagged version of extdebtGDP, Stata shows the following error:

  xtreg rgdpgrowth sxp sxp2 extdebtGDP L.extdebtGDP i.year, fe vce(robust)
no observations
r(2000);

In a previous related post, I saw Dr. Nick Cox mentioning that string variables/missing observations may cause this error.

 count if !missing( rgdpgrowth, sxp, sxp2, extdebtGDP)
  315

. count if !missing( rgdpgrowth, sxp, sxp2, extdebtGDP, L.extdebtGDP)
    0

xtdes

 country:  1, 2, ..., 104                                    n =         65
    year:  1975, 1980, ..., 1995                             T =          5
           Delta(year) = 1 unit
           Span(year)  = 21 periods
           (country*year uniquely identifies each observation)

Distribution of T_i:   min      5%     25%       50%       75%     95%     max
                         3       4       5         5         5       5       5

     Freq.  Percent    Cum. |  Pattern*
 ---------------------------+----------
       60     92.31   92.31 |  11111
        3      4.62   96.92 |  ..111
        2      3.08  100.00 |  .1111
 ---------------------------+----------
       65    100.00         |  XXXXX
 --------------------------------------
 *Each column represents 5 periods.

Any help will be highly appreciated. Thanks.

[Nick Cox]

xtset would not have worked at all with a string identifier. xtset did work, but you forgot to specify delta(5) to indicate 5 year spacing.

[Taz Raihan]

Amazing! Can't thank you enough!

[Carlo Lazzaro]

Taz:
as a sidelight (and despite that squared exp does not seem to play any role in your regression model), relying on -fvvarlist- notation for creating catagorical variables and interactions, is still then best approach (see also the relationships between -fvvralist-, -margins- and -marginsplot-).
Hence, your second code could have been written as:

 xtreg rgdpgrowth c.sxp##c.sxp extdebtGDP L.extdebtGDP i.year, fe vce(robust)

[Taz Raihan]

Carlo Lazzaro thank you for chiming in. I deliberately did not use c.sxp##c.sxp since I was using outreg2 to output my results in tex format with variable names replaced by their labels. If I had used c.sxp##c. sxp, then I would have to manually change the labels of squared sxp in each of the tex files that were being generated. But is my method going to affect the results in a significant way?

Again, thanks for your advice.

[Carlo Lazzaro]

Taz:
Admittedly, I did not notice that you mentioned -outreg2- in your previous post.
No, your results will be the same; set aside the post-estimation issues (see -margins-and -marginsplot-), the only difference is that plugging in a squared term by hand makes Stata think that the linear and the squared term refer to two different predictors, whereas it obviously not the case.

[Taz Raihan]

Carlo Lazzaro thank you for enlightening. Learning new stuff every moment.

[Jur Hoekstra]

Hello everybody,

I am new to stata and up till now previous posts helped me to where I am now, thanks for that to all.
I have ran a rangestat regression to get the coefficient for a four factor model regression, I think I succeeded in this. For this I used the following regression

rangestat (reg) ER ERM SMB HML WML , interval(Date . -1)

The problem however is that when I want to use the coefficient in another regression: xtreg, this returns the error message: r2000 no observations.
This regression I mentioned does not have string variables, but does have missing values so I think the panel data is unbalanced.

For the first one mentioned below there is a table that is produced, but with a zero as the coefficient in the results, for the second one not even a table can be produced mentioning the error r2000.

xtreg b_consNet Ratingnr i.Date
outreg2 using 1.1A1.doc, replace ctitle (1) tstat addtext (Firm FE, YES, Time FE, YES)

xtreg b_consNet SRI EstimatedFundLevelNetFlow Averagetimereturn Tenurelongest Age Turnover NETER Holdings Stdev Load dummy90 SIZE_ln i.Date
outreg2 using 1.2A.doc, replace ctitle (1) tstat addtext (Firm FE, YES, Time FE, YES)

I really do not know where I go wrong. Do you guys have any tips or pointers that can lead me in the right direction?

Thanks in advance,
Jur

[Sowmya Kannan]

Hi, I am trying to run a panel data regression and I am facing the issue of "no observations" error. My dependent variable is continuous that is an acquisition deal level variable. SDCDealNo uniquely identifies each deal. Independent variable is binary that is a country level variable. Both are not string. I checked for missing values.

count if !missing (ownership p_shift)
missing not found
r(111);

xtreg ownership l.p_shift
no observations
r(2000);

This is how I set my panel data.

xtset SDCDealNo year
       panel variable:  SDCDealNo (unbalanced)
        time variable:  year, 1995 to 2020
                delta:  1 unit

xtdes

SDCDealNo:  1.669e+08, 1.800e+08, ..., 4.067e+09            n =      46309
    year:  1995, 1996, ..., 2020                             T =         26
           Delta(year) = 1 unit
           Span(year)  = 26 periods
           (SDCDealNo1*year uniquely identifies each observation)

Distribution of T_i:   min      5%     25%       50%       75%     95%     max
                         1       1       1         1         1       1       1

     Freq.  Percent    Cum. |  Pattern
 ---------------------------+----------------------------
     2313      4.99    4.99 |  ...1......................
     2300      4.97    9.96 |  .....1....................
     2109      4.55   14.52 |  ............1.............
     2086      4.50   19.02 |  ........................1.
     2085      4.50   23.52 |  ....1.....................
     2082      4.50   28.02 |  .......................1..
     2071      4.47   32.49 |  .........................1
     1990      4.30   36.79 |  ....................1.....
     1984      4.28   41.07 |  ......................1...
    27289     58.93  100.00 | (other patterns)
 ---------------------------+----------------------------
    46309    100.00         |  XXXXXXXXXXXXXXXXXXXXXXXXXX

Any help much appreciated!

[Carlo Lazzaro]

Sowmya:
you're probably experiencing something like:

use "C:\Program Files\Stata18\ado\base\a\auto.dta"
(1978 automobile data)

. g id=_n

. g time=1

. tsset id time

Panel variable: id (strongly balanced)
 Time variable: time, 1 to 1
         Delta: 1 unit

. regress price l.mpg
no observations
r(2000);

.

[Daniel Feenberg]

To run a panel regression you would need some units with observations in multiple years. It appears that you have only one observation for each SDCDealno.