how to randomly sample from an unbalanced dataset

[Raymon Lucas]

Hey colleages,

I would randomly sample 1 observation from the following dataset:

identifier number of obs.
1 1004
2 321
3 5
4 1
5 16
6 114
.....

I had expected to sample 1 obs. for each identifier, but I soon notice that the data points are very unevenly distributed.

I would like to know how I should sample under such circumstances. Note: I sample in order to construct a negative case dataset for my research.

Thanks a lot!

Raymon Lucas

[Carlo Lazzaro]

Raymon:
you may want to try:

sample 1, count by(identifier)

As expected, the -id- with one observation only will return one and only one value, as you can see from the following toy-example:

. set ob 11
Number of observations (_N) was 0, now 11.

. g id=1 in 1/10


. replace id=2 in 11


. g A=runiform()

. list

     +---------------+
     | id          A |
     |---------------|
  1. |  1   .8712187 |
  2. |  1   .7664683 |
  3. |  1   .2512555 |
  4. |  1   .1663648 |
  5. |  1   .7437958 |
     |---------------|
  6. |  1   .9805113 |
  7. |  1   .7295772 |
  8. |  1   .9011049 |
  9. |  1   .2643649 |
 10. |  1   .8856509 |
     |---------------|
 11. |  2    .882112 |
     +---------------+

. sample 1, count by(id)
(9 observations deleted)

. list

     +---------------+
     | id          A |
     |---------------|
  1. |  1   .9805113 |
  2. |  2    .882112 |
     +---------------+

.

Your problem, if any, has nothing to do with the random sampling approach, but it is data related (and I fear there's nothing you can do about that).

[Raymon Lucas]

Carlo, many thanks for your helpful reply, which makes me reassured about my choice. Thank you again! Raymon