Displaying number of observations after "Repest" command

[Mor Zahavi]

I use the user written command "Repest" to analyse PISA survey data.
I then use eststo and esttab to export the results to LaTeX.
"Repest" stores the number of observation and r-squared as coefficients and not as statistic, so I have to use estadd to add them as a scalar.
My problem is that when I use estadd I get an error message.


The code I use is -

repest PISA, estimate(stata: reg pv@math female) ///
results(add(N r2)) fast
estadd scalar No = e_N
estadd scalar r2 = e_r2
eststo m1
esttab m1

When I run esttab I get

----------------------------
                      (1)   
                            
----------------------------
female             -13.62***
                 (-12.30)   

_cons               456.4***
                 (411.44)   

e_N                515956   
                      (.)   

e_r2              0.00422***
                   (5.99)   
----------------------------
N                           
----------------------------

But the code gives an error message after the command

estadd scalar No = e_N

The error message is:

"e_N not found"

Could you help me extract these scalars (N and r2) after the "Repest" command?

[William Lisowski]

Unfortunately, neither your post nor the documentation in the output of help repest shows a set of commands that would allow me to test my guess on actual data. With that said, perhaps your code should be

// the results option seems intented to instruct repest to retain
// the e(N) and e(r2) from the regress command in the repest estimation results
repest PISA, estimate(stata: reg pv@math female) ///
    results(add(N r2)) fast
eststo m1
esttab m1, r2

[Mor Zahavi]

Hi William, I have created a subsample of the data for you to test commands. What would be the best way to to share it with you?

[William Lisowski]

Did you first try my suggested code in post #2? I think the estadd commands are unnecessary. The results() option on the repest command should add e(N) and e(r2) from the regress command to the set of results that repest leaves in e(), After running the repest command, run the command ereturn list to see what is now in e().

[Mor Zahavi]

Yes I did. It did not work. When I run ereturn list after the repest command I get the following:
end of do-file

. ereturn list

scalars:
                  e(R) =  166.289

macros:
                e(cmd) : "."
         e(properties) : "b V"

matrices:
                  e(b) :  1 x 163
                  e(V) :  163 x 163

(PS: How do you make a text in your reply appear as code in line, without it being surrounded in grey and moved to a new paragraph?)

[William Lisowski]

I think I was misled by the documentation provided in the help repast output. (Note: I made that look like code without being a code block by changing the font to Courier New using the Font dropdown in the "Advanced Editing" toolbar which if you don't see it, is revealed by clicking on the A in the middle of the three icons on the right in the posting window.)

Your example in post #5, that the "estimates results" returned in e() is not where repest is saving the "results" specified the results() option. Or at least, they are not stored in separate scalars.

If they are actually stored as coefficients, as you suggested, then they would be elements in the returned matrix e(b). Perhaps adapting the following example to your situation will help with your problem. If it does, I'm sorry for having initially set out on the wrong path.

. sysuse auto, clear
(1978 Automobile Data)

. quietly regress price length weight

. matrix list e(b)

e(b)[1,3]
        length      weight       _cons
y1  -97.960312   4.6990649   10386.541

. matrix coeff = e(b)

. // retrieve using row and column numbers
. scalar b_n = coeff[1,2]

. // retrieve using row and column names
. scalar b_v = coeff[rownumb(coeff,"y1"),colnumb(coeff,"weight")]

. display b_n " " b_v
4.6990649 4.6990649

[Mor Zahavi]

Thanks William,
Your thorough answer has cleared everything. I can now extract the scalars and use them in my tables. It was great that you have bothered to provide the command for the column names, since I run 3 regressions each time, and the number of the relevant column changes each time (since each regression has more variables).
Thanks a bunch!

[Mor Zahavi]

Dear William, I would like to kindly ask for your assistance again on a similar matter -

mixed numscore1 lbe female books* immig* ,|| schoolid:

and get the following output:

Performing EM optimization: 

Performing gradient-based optimization: 

Iteration 0:   log likelihood = -661767.19  
Iteration 1:   log likelihood = -661767.19  

Computing standard errors:

Mixed-effects ML regression                     Number of obs     =    492,472
Group variable: schoolid                        Number of groups  =      1,535

                                                Obs per group:
                                                              min =          1
                                                              avg =      320.8
                                                              max =      2,656

                                                Wald chi2(11)     =   55475.89
Log likelihood = -661767.19                     Prob > chi2       =     0.0000

------------------------------------------------------------------------------
   numscore1 |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         lbe |  -.1191841   .0028493   -41.83   0.000    -.1247686   -.1135997
      female |  -.1540483   .0026483   -58.17   0.000    -.1592389   -.1488577
      books1 |  -.8672732   .0061195  -141.72   0.000    -.8792671   -.8552792
      books2 |  -.6698036   .0059917  -111.79   0.000     -.681547   -.6580601
      books3 |  -.3961525    .005769   -68.67   0.000    -.4074596   -.3848455
      books4 |  -.1850617   .0062339   -29.69   0.000      -.19728   -.1728434
      books5 |   .0388173   .0065941     5.89   0.000     .0258932    .0517415
      books6 |          0  (omitted)
      immig1 |   .0275492   .0111689     2.47   0.014     .0056586    .0494397
      immig2 |   .0598328   .0113963     5.25   0.000     .0374966    .0821691
    immig_mo |  -.0616333   .0066826    -9.22   0.000     -.074731   -.0485356
    immig_fa |  -.0567575   .0071256    -7.97   0.000    -.0707234   -.0427917
       _cons |   .6041496   .0100567    60.07   0.000     .5844388    .6238603
------------------------------------------------------------------------------

------------------------------------------------------------------------------
  Random-effects Parameters  |   Estimate   Std. Err.     [95% Conf. Interval]
-----------------------------+------------------------------------------------
schoolid: Identity           |
                  var(_cons) |   .0952685   .0043435      .0871247    .1041735
-----------------------------+------------------------------------------------
               var(Residual) |   .8534137    .001723      .8500434    .8567974
------------------------------------------------------------------------------
LR test vs. linear model: chibar2(01) = 8925.74       Prob >= chibar2 = 0.0000

How do extract the details on the number of groups and the min/max number of observation in a group?