sureg producing different results every time I run program

[Lucas Rengifo-Keller]

Hi all,

I am wanting to estimate seemingly unrelated regressions (SUR). I know of two commands that are designed for such estimation: sureg and xtsur, where sureg is designed for balanced panels and xtsur fit for unbalanced panels. Please let me know if I'm incorrect here.

I am working with an unbalanced panel dataset that looks at different product prices (chocolate, chips, and candy). I tried balancing it by keeping only the dates where chocolate!=. & chips!=. & candy !=. . In other words, the dates where prices for all three products are observed. Then comes my problem: every time I run my code, the sureg parameters are different.


I should note that I only have one dependent variable (price). I separate price, by(product) and this lets me run the code, but again the results are different each time I hit 'do.' Any suggestions as to why this may be?


P.S. I have tried using xtsur with some success, but I am required to average all my series at the daily level in order to have dates that don't repeat (and xtset the data). Averaging obviously greatly reduces my observation count, which I do not want.

[Lucas Rengifo-Keller]

See what my dependent variable data look like below. These are individual product transactions that I then merge with my independent variables (income per capita in the region, shipping rates, etc). I have independent variables that are defined for each product (e.g., income_candy, income_chocolate, income_chips) that I regress on.

So my code looks more or less like this: sureg (price_candy log_time_candy income_candy bulk_purchase_candy shipping_rates_candy) (price_chocolate log_time_chocolate income_chocolate bulk_purchase_chocolate ) (price_chips log_time_chips income_chips bulk_purchase_chips shipping_rates_chips)

* Example generated by -dataex-. For more info, type help dataex
clear
input float(price_candy price_chocolate price_chips price_date)
  2.995732   4.892852 -1.2039728 19360
 2.0794415  1.2039728   .4054651 19361
 3.6888795   5.703783  -.4054651 19389
  .9162908   6.838762   .6931472 19398
  2.995732  4.4228487          0 19412
  3.178054   6.214608  1.7917595 19444
 1.7917595   .8675005   .6931472 19446
  .9162908   5.298317   .2876821 19454
  .6931472  3.3242364  -.9808292 19479
 1.0986123  2.4079456          0 19484
  .3364722   1.609438   .6931472 19535
  1.609438  2.4079456          0 19549
  2.014903   2.631089  1.3862944 19584
  2.995732   2.995732          0 19590
 2.0794415   5.298317  -.4054651 19593
         0  1.7837913   .4054651 19649
 1.0986123   1.966113  .06062462 19652
  1.609438   5.298317   .6931472 19658
         0  3.3242364 -.18232156 19666
  1.609438  2.4079456 -.51082563 19678
 2.3025851    6.50229  1.3862944 19687
  .6931472   6.907755  -.6931472 19753
   1.94591  2.3025851          0 19776
  2.484907  2.1202636   .1541507 19790
  .6931472  4.7105308          0 19794
  .6931472    6.55108  -1.609438 19795
 1.3862944   3.170086          0 19796
  3.465736   2.748872  1.3862944 19801
 3.6888795   2.748872          0 19803
  .9162908  2.1892564  1.3862944 19810
  1.609438    6.50229  -.9162908 19813
 1.7917595   5.298317          0 19821
 1.7917595   5.585999          0 19824
 1.3862944   .6751287  -.6931472 19827
 3.6888795  4.6051702  1.3862944 19829
  1.609438   2.525729          0 19837
  .6931472   5.585999          0 19841
  1.609438   6.279147   .6931472 19842
  1.609438   3.660709 -.22314355 19845
  2.525729  2.1202636  .18232156 19849
 1.7917595   .6931472          0 19851
  2.995732   3.912023 -.22314355 19879
 1.0986123   3.506558  -.6931472 19882
 2.3025851   5.809143  1.6486586 19883
 1.3862944   3.912023          0 19886
  2.944439   6.907755  -.2876821 19891
  .9162908   .1743534   .2876821 19893
         0  4.6051702  1.3862944 19894
  1.252763   .1743534  1.3862944 19918
 1.0986123  4.6051702  1.1631508 19919
 1.0986123   6.279147          0 19924
 1.0986123  3.3242364          0 19928
  .6931472   5.298317          0 19951
  .6931472 -4.4308167          0 20016
         0   6.907755   .6931472 20024
 2.3025851    1.89712  -.6931472 20041
  1.609438  1.8325815 -.51082563 20092
 1.0986123  4.1351666  -.4054651 20184
 1.3862944  1.3664917   .2876821 20205
 1.0986123   3.218876  -.4054651 20227
 2.3025851  1.5896353  1.2669476 20229
 3.3322046   3.624341 -.18232156 20241
  .6931472  1.7837913          0 20245
  .6931472  3.8066626  -.6931472 20247
 1.7917595   3.506558          0 20248
  .6931472   2.525729  -.2876821 20280
 4.6051702   2.525729  -.3566749 20285
  .6931472  2.2827826  -.4054651 20302
  .9162908  1.7837913          0 20306
  2.995732   3.218876          0 20318
 1.3862944  4.1351666 -1.0498221 20320
         0  1.7837913          0 20327
 2.3025851  -.9242589   .1541507 20343
 2.3025851   4.317488  -.9162908 20344
  1.609438   2.525729   .8754687 20365
 1.7917595   1.609438  1.7917595 20387
  1.609438   1.654174   .6931472 20388
   1.94591  2.2827826          0 20389
 .10436001  .46203545 -.10536052 20396
 1.0986123   1.482686          0 20398
 1.3862944   3.085344  -.9162908 20399
  1.252763  1.9261075 -.18232156 20402
  2.995732  1.0788096 -.22314355 20405
 1.3862944   1.609438  -2.995732 20409
  .9162908  1.7837913          0 20439
 1.3862944   5.521461  -1.609438 20470
  .6931472  4.1351666          0 20471
  2.014903   2.253795          0 20503
  .4054651   3.912023  -1.609438 20511
  .9162908   3.170086   .2876821 20519
         0  1.5896353  -.6931472 20520
  .9162908   4.787492 -.22314355 20526
  1.609438   3.218876  -.6931472 20529
  1.609438  4.6051702  -5.521461 20542
 1.0986123  1.5896353  -.2876821 20549
 1.3862944  2.2827826 -.22314355 20553
  .6931472  4.6051702   .2876821 20564
  .6931472   4.199705 -.51082563 20568
 .22314355  2.2827826          0 20570
-.22314355  3.6888795   .2876821 20572
end
format %tdnn/dd/CCYY price_date

[Carlo Lazzaro]

Lucas:
your data example does not allow to replicate your equations.
In addition, while the comminity-contributed module (as you're kindly requested to mention it for good reasons detailed in the FAQ) -xtsur- was actually developed for dealing with unbalanced panel datasets, -sureg- was oroginally developed for regression equations sharing the same regerssand.