Difficulty understanding the difference using bysort with parenthesis

[Jose Rosa]

Hi all,

I have been trying to look at different forums, and found similar answers, but I have a certain application I cannot find an answer to.


I have sorted my data by a group variable and want to keep a particular group if an indicator variable denoted indicatorvar ==1 at least once in the group variable. However, I get two different sized datasets when doing the following two different methods, holding other code equal:

bysort group: keep if indicatorvar[_N]

bysort (indicatorvar): keep if indicatorvar[_N]

Upon browsing both results, it seems as if it keeps the entire group as long as the indicator variable is activated in at least one observation within a group. However, the first result is much smaller observation-wise than the second result.

I cannot understand why it is doing it. Any help would be appreciated.


Similar forum is here:-by- syntax of adding another variable in brackets - Statalist

[Øyvind Snilsberg]

bysort group: keep if indicatorvar[_N]

keeps groups whose last value of indicatorvar is 1 given the current sort order

bysort group (indicatorvar): keep if indicatorvar[_N]

keeps groups whose largest value of indicatorvar is 1

* Example generated by -dataex-. For more info, type help dataex
clear
input byte(group indicatorvar)
1 0
1 0
1 1
2 0
2 1
2 0
end

bysort group: g a = indicatorvar[_N]
bysort group (indicatorvar): g b = indicatorvar[_N]

list ,sep(6)

     +--------------------------+
     | group   indica~r   a   b |
     |--------------------------|
  1. |     1          0   1   1 |
  2. |     1          0   1   1 |
  3. |     1          1   1   1 |
  4. |     2          0   0   1 |
  5. |     2          0   0   1 |
  6. |     2          1   0   1 |
     +--------------------------+

[Jose Rosa]

bysort group: keep if indicatorvar[_N]

keeps groups whose last value of indicatorvar is 1 given the current sort order

bysort group (indicatorvar): keep if indicatorvar[_N]

keeps groups whose largest value of indicatorvar is 1

* Example generated by -dataex-. For more info, type help dataex
clear
input byte(group indicatorvar)
1 0
1 0
1 1
2 0
2 1
2 0
end

bysort group: g a = indicatorvar[_N]
bysort group (indicatorvar): g b = indicatorvar[_N]

list ,sep(6)

+--------------------------+
| group indica~r a b |
|--------------------------|
1. | 1 0 1 1 |
2. | 1 0 1 1 |
3. | 1 1 1 1 |
4. | 2 0 0 1 |
5. | 2 0 0 1 |
6. | 2 1 0 1 |
+--------------------------+

Thank you so much, extremely effective answer!