Estadd does not reproduce scalar saved post resetxt

[Stella Bagasheva]

Hello Statalists,

I am running some random effects regressions on panel data and would like to test which specification is correct using the Ramsey test. I am using the user-created command resetxt downloaded from SSC.
Afterwards, I would like to organise my results in a table format to compare them using the user-created estadd.

However, estadd does not reproduce the statistics in the table. Would anyone be able to help with this? Please see my data and code below.

* Example generated by -dataex-. To install: ssc install dataex
clear
input float(id year cost input input_squared)
10 2017 12.439766  7.387097  54.56921
10 2018 13.060278  7.338101  53.84773
10 2019 12.811565  7.595946  57.69839
11 2017   11.9424  6.868429  47.17531
11 2018 12.617618   6.82206   46.5405
11 2019 12.263633  6.853036  46.96411
12 2017  9.337701  2.475606  6.128623
12 2018  9.926363  2.429237  5.901192
12 2019  9.417656 2.4602134   6.05265
13 2017  8.440928  .9367855   .877567
13 2018 10.189826  .8904165  .7928416
13 2019  9.953368   .921393  .8489651
14 2017  9.615168  2.876588   8.27476
14 2018 10.391204 2.8302195  8.010142
14 2019 11.316359  2.861196  8.186442
15 2017  12.19255 4.3916397   19.2865
15 2018  12.25161  4.345271  18.88138
15 2019 12.307883 4.3762474  19.15154
16 2017  9.043903 2.8293495 8.0052185
16 2018  9.867072  2.782981  7.744981
16 2019   9.33295  2.813957  7.918355
17 2017   8.55707  2.485013  6.175291
17 2018  9.464444 2.4386444  5.946987
17 2019  8.755593  2.469621  6.099028
18 2017  14.39939  8.888922  79.01293
18 2018 14.115676    8.8381  78.11202
18 2019 14.000032  8.839617  78.13882
19 2017  10.84751  5.631671  31.71572
19 2018 11.298195  5.585302   31.1956
19 2019 10.978483  5.616279 31.542585
20 2017 14.017168  8.601939 73.993355
20 2018 13.791075  8.450449  71.41009
20 2019 13.897657  8.447175  71.35477
end

global reg1 cost input
global reg2 cost input input_squred

    forvalues i = 1(1)2{
    xtset name year
    
    *Run Pooled random effect regressions with clustered standard errors at id level and store results.
    eststo re`i': qui xtreg ${reg`i'} , re vce(cluster id)
    sort id year
        
    *RESET test
    resetxt ${reg`i'} , model(xtre) id(id) it(year)
    estadd scalar RESET_Ramsay_stata = e(resetf3p): re`i'
    estadd scalar RESETL_DG_stata = e(resetl3p): re`i'
    xtset id year
    
    }    

estout re1 re2 , ///
        cells(b(star fmt(3)) se(par fmt(3)))      ///
        stats(RESET_Ramsey_stata RESETL_DG_stata ///
        ,fmt(3)) starlevels(* 0.10 ** 0.05 *** 0.01)

Here is the estadd output, which does not reproduce the test statistic:

                               re1             re2  
                             b/se            b/se  
----------------------------------------------------
input                     0.470***        0.263***
                            (0.018)         (0.073)  
input_squared                               0.037***
                                          (0.007)  
constant                    6.587***        3.502***
                          (0.116)         (0.190)  
----------------------------------------------------
RESET_Ramsey_stata                                  
RESETL_DG_stata                 .               .

[Andrew Musau]

See https://www.statalist.org/forums/for...alar-pseudo-r2

[Stella Bagasheva]

Thanks Andrew. I tried saving the scalars using estout as per your suggestion and then using estadd but unfortunately it did not work - there is still a . where the number should be. Do you have any other suggestions? Might it not work because I am running a loop and the scalar is not saved properly?

[Andrew Musau]

Try

*RESET test
    resetxt ${reg`i'} , model(xtre) id(id) it(year)
    estadd scalar RESET_Ramsay_stata = `=e(resetf3p)': re`i'
    estadd scalar RESETL_DG_stata = `=e(resetl3p)': re`i'
    xtset id year

[Stella Bagasheva]

YES, IT WORKS!

Thanks so much, Andrew