runiform function and r(N) scalar

[Nir Godin]

Hi,

I am trying to use the runiform function to randomally select 1/3 of the observations. Unfortunately, it seems that the use of the function affects unrelated results.
For example, look at this picture:

As you can see, I am writing the exact same sum command in lines 19 and 25. These are the results of the return list:

The return lists are totally different, even though this is the same sum command, and even though the ruinform command is not related in any way to the summed variable.
Does anyone knows what may causes this?

Thanks,
Nir.

[Carlo Lazzaro]

Nir:
Stata returns this seemingly weird results because your second -sum- includes -RandomNumber-, too, as you can see from the following toy-example:

. use "C:\Program Files\Stata16\ado\base\a\auto.dta"
(1978 Automobile Data)


. sum if rep78==3 [fw=weight]

    Variable |        Obs        Mean    Std. Dev.       Min        Max
-------------+---------------------------------------------------------
        make |          0
       price |     98,970    7003.154    3748.311       3291      15906
         mpg |     98,970    18.70183    4.011309         12         29
       rep78 |     98,970           3           0          3          3
    headroom |     98,970     3.20774    .7557963          2        4.5
-------------+---------------------------------------------------------
       trunk |     98,970    15.72224    3.516451          9         23
      weight |     98,970    3464.887     716.731       1830       4840
      length |     98,970    198.1066    19.35086        142        233
        turn |     98,970    41.80277    4.130643         31         51
displacement |     98,970    248.2548    91.34204         79        425
-------------+---------------------------------------------------------
  gear_ratio |     98,970    2.797555    .3843342       2.24       3.72
     foreign |     98,970    .0609276    .2391986          0          1

. return list

scalars:
                  r(N) =  98970
              r(sum_w) =  98970
               r(mean) =  .0609275538041831
                r(Var) =  .0572159651058491
                 r(sd) =  .2391985892639191
                r(min) =  0
                r(max) =  1
                r(sum) =  6030

. gen RND=runiform() if price>3000 & price!=.

. sum if rep78==3 [fw=weight]

    Variable |        Obs        Mean    Std. Dev.       Min        Max
-------------+---------------------------------------------------------
        make |          0
       price |     98,970    7003.154    3748.311       3291      15906
         mpg |     98,970    18.70183    4.011309         12         29
       rep78 |     98,970           3           0          3          3
    headroom |     98,970     3.20774    .7557963          2        4.5
-------------+---------------------------------------------------------
       trunk |     98,970    15.72224    3.516451          9         23
      weight |     98,970    3464.887     716.731       1830       4840
      length |     98,970    198.1066    19.35086        142        233
        turn |     98,970    41.80277    4.130643         31         51
displacement |     98,970    248.2548    91.34204         79        425
-------------+---------------------------------------------------------
  gear_ratio |     98,970    2.797555    .3843342       2.24       3.72
     foreign |     98,970    .0609276    .2391986          0          1
         RND |     98,970    .5004157    .2550064   .1649395   .9881987

. return list

scalars:
                  r(N) =  98970
              r(sum_w) =  98970
               r(mean) =  .5004157180664237
                r(Var) =  .0650282703071466
                 r(sd) =  .2550064122863317
                r(min) =  .1649395227432251
                r(max) =  .9881986975669861
                r(sum) =  49526.14361703396

.