Reshape only for certain responses

[Anne Todd]

Hi all, with the data I have below, only some respondents provided answers for a second friend. As of now, I am reshaping this data to be long and then trying to remove extraneous cases, i.e., when someone only responded for one friend, because they still get a second row when I reshape for the full dataset. Is there a way to essentially do this within the reshape itself?


Here is what I am currently doing (in case it's relevant, in the full dataset there are something like 30 variables that are included in the reshape:

reshape long num_drinks_friend yr_attended_friend, i(id) j(friend)
drop if friend == 2 & multiple_friend == 0 // If they do not respond for multiple friends but have a 2nd friend observation after the reshape

Some data:

* Example generated by -dataex-. For more info, type help dataex
clear
input float(id multiple_friends num_drinks_friend1 num_drinks_friend2 yr_attended_friend1 yr_attended_friend2 race_white)
 1 0 3 .    .    . 1
 2 0 3 . 2007    . 1
 3 0 3 . 2010    . 0
 4 0 4 . 2011    . 1
 5 1 5 3 1992 1984 0
 6 0 3 . 2007    . 1
 7 0 3 . 2008    . 1
 8 0 3 . 2012    . 1
 9 1 3 6 2011    . 0
10 1 4 3 2008 2006 0
11 0 3 . 2011    . 1
12 0 4 . 2007    . 1
13 1 6 3 2012 1981 0
14 0 3 . 2008    . 1
15 0 . . 2009    . 1
16 1 5 5    .    . 0
17 1 3 5 1992    . 0
18 0 5 . 2010    . 1
19 0 3 . 2008    . 1
20 0 3 . 2008    . 0
end

[Clyde Schechter]

No, you can't modify the -reshape- to handle this. The approach you have taken, of dropping the observations that are uninformative, is the way to go.