Time-in treatment for panel data

[BJ Cunningham]

Hi - I am trying to create a time-varying variable that identifies how long a person has been receiving intervention (i.e., time between assessment 1 and assessment 2; time between assessment 1 and assessment 3, time between assessment 1 and assessment 4, etc.)

ID	Score	Assessment	Age (months)	Time receiving intervention
1	20	1	18	0
2	30	1	20	0
2	35	2	26	*Should be 6 (26-20 mos.)
3	15	1	24	0
4	7	1	30	0
5	10	1	26	0
5	16	2	30	*Should be 4 (30-26 mos.)
5	18	3	37	*Should be 11 (37-26 mos.)
6	40	4	42	*Should be 16 (42-26 mos.)

Thank you so much in advance for your ideas!

[Carlo Lazzaro]

BJ:
assuming you have no missing data, you can try:

bysort id: g A=agemonths[_n]- agemonths[_n-1]
replace A=0 if A==.
bysort id: g B=sum(A)

[BJ Cunningham]

Thank you Carlo! I tried the code you sent, and it works really well for calculating the time between assessment 1 and assessment 2, but it doesn't calculate the time between assessment 1 and assessment 3. Instead it gives me the difference between assessment 1 and assessment 2, then the difference between assessment 2 and assessment 3. Thanks so much for any other ideas you may have! BJ

[Nick Cox]

BJ: Please note the advice on how to present data in http://www.statalist.org/forums/help#stata

and then consider the following. Your last observation looks wrong to me in some detail or another, probably that the identifier should really be 5.

clear
input ID    Score    Assessment    Age  Wanted
1    20    1    18    0    
2    30    1    20    0    
2    35    2    26    6
3    15    1    24    0    
4    7    1    30    0    
5    10    1    26    0    
5    16    2    30    4
5    18    3    37    11
6    40    4    42    16
end

bysort ID : gen Suggested = Age - Age[1]

list, sepby(ID)

     +-------------------------------------------------+
     | ID   Score   Assess~t   Age   Wanted   Sugges~d |
     |-------------------------------------------------|
  1. |  1      20          1    18        0          0 |
     |-------------------------------------------------|
  2. |  2      30          1    20        0          0 |
  3. |  2      35          2    26        6          6 |
     |-------------------------------------------------|
  4. |  3      15          1    24        0          0 |
     |-------------------------------------------------|
  5. |  4       7          1    30        0          0 |
     |-------------------------------------------------|
  6. |  5      10          1    26        0          0 |
  7. |  5      16          2    30        4          4 |
  8. |  5      18          3    37       11         11 |
     |-------------------------------------------------|
  9. |  6      40          4    42       16          0 |
     +-------------------------------------------------+

[BJ Cunningham]

This works perfectly. Thank you so much! BJ