generate variable considering interuptions

[Celine St]

Hello
I am currently working with Panel Data. I have one variable which tells me, if a person lives in poverty (dummy variable). Since I am interessted in long-term poverty, I need a new variable, which tells me how long the poverty spell lastet (or the longest if several poverty spells) (without interuption).

I tried:

by idpers, sort: egen firsttime = min(cond(poverty==1, year, .))
by idpers: egen lasttime = max(cond(poverty==1, year, .))


but with this, it doesn't consider interuptions, but rather givs me the first and last time poverty occurred to the same person.

Thank you very much for the help.

[Hemanshu Kumar]

Here's one way:

* ignore this top bit; it is just used to simulate some data
clear
set obs 20
set seed 111
gen byte idpers = ceil(_n/10)
bys idpers: gen int year = 2000 + _n
gen poverty = runiformint(0,1)

* start your code from here
sort idpers year
by idpers: gen byte spell_start = poverty == 1 & (_n == 1 | poverty[_n-1] == 0)
by idpers: gen spell_id = cond(poverty == 1, sum(spell_start), .)
bys idpers spell_id: egen spell_length = count(spell_id)
by idpers: egen wanted = max(spell_length)
drop spell_start spell_id spell_length
sort idpers year

which produces:

. li, sepby(idpers) noobs
  +----------------------------------+
  | idpers   year   poverty   wanted |
  |----------------------------------|
  |      1   2001         1        6 |
  |      1   2002         1        6 |
  |      1   2003         1        6 |
  |      1   2004         1        6 |
  |      1   2005         1        6 |
  |      1   2006         1        6 |
  |      1   2007         0        6 |
  |      1   2008         0        6 |
  |      1   2009         0        6 |
  |      1   2010         1        6 |
  |----------------------------------|
  |      2   2001         0        3 |
  |      2   2002         1        3 |
  |      2   2003         1        3 |
  |      2   2004         1        3 |
  |      2   2005         0        3 |
  |      2   2006         1        3 |
  |      2   2007         1        3 |
  |      2   2008         0        3 |
  |      2   2009         0        3 |
  |      2   2010         1        3 |
  +----------------------------------+

[Nick Cox]

There is a Stata literature and there are programs to help. On the latter, see e.g. tsspell from SSC.

* Example generated by -dataex-. For more info, type help dataex
clear
input float(id year poverty)
1 2018 0
1 2019 0
1 2020 1
1 2021 1
1 2022 1
2 2018 1
2 2019 1
2 2020 0
2 2021 0
2 2022 1
end

tsset id year 

tsspell, cond(poverty == 1)
by id : egen longest = max(_seq)
by id : egen nspells = total(_end)

list, sepby(id) 


     +----------------------------------------------------------------+
     | id   year   poverty   _seq   _spell   _end   longest   nspells |
     |----------------------------------------------------------------|
  1. |  1   2018         0      0        0      0         3         1 |
  2. |  1   2019         0      0        0      0         3         1 |
  3. |  1   2020         1      1        1      0         3         1 |
  4. |  1   2021         1      2        1      0         3         1 |
  5. |  1   2022         1      3        1      1         3         1 |
     |----------------------------------------------------------------|
  6. |  2   2018         1      1        1      0         2         2 |
  7. |  2   2019         1      2        1      1         2         2 |
  8. |  2   2020         0      0        0      0         2         2 |
  9. |  2   2021         0      0        0      0         2         2 |
 10. |  2   2022         1      1        2      1         2         2 |
      +----------------------------------------------------------------+

On the former, here are some suggestions.

SJ-18-1 dm0078_3 . . . . . . . . . . . . . . . . Software update for newspell
(help newspell if installed) . . . . . . . . . . . . . . . H. Kroger
Q1/18 SJ 18(1):290
newspell combine now combines overlapping spells correctly

SJ-17-2 dm0078_2 . . . . . . . . . . . . . . . . Software update for newspell
(help newspell if installed) . . . . . . . . . . . . . . . H. Kroger
Q2/17 SJ 17(2):515--516
newspell fillin now fills in gaps correctly; newspell gap bug
fixed for finding gaps

SJ-16-1 dm0078_1 . . . . . . . . . . . . . . . . Software update for newspell
(help newspell if installed) . . . . . . . . . . . . . . . H. Kroger
Q1/16 SJ 16(1):244
bug fixed for combining two types of spells and dataset is now
sorted as specified in sort()

SJ-15-1 dm0078 . . . . . . . newspell: Easy management of complex spell data
(help newspell if installed) . . . . . . . . . . . . . . . H. Kroger
Q1/15 SJ 15(1):155--172
introduces a command for merging or cutting off spells,
finding gaps in the data, filling those gaps, transforming
spell data to wide or long format, and creating completely
new states from overlapping spells

SJ-15-1 dm0079 . . . . . . . . . . . . . . . Stata tip 123: Spell boundaries
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . N. J. Cox
Q1/15 SJ 15(1):319--323 (no commands)
shows how to identify spells


SJ-7-2 dm0029 . . . . . . . . . . . . . . Speaking Stata: Identifying spells
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . N. J. Cox
Q2/07 SJ 7(2):249--265 (no commands)
shows how to handle spells with complete control over
spell specification

[Celine St]

Thank you both! I got it and it work perfectly