I took a rather different tack from Nick C. I explicitly looped over
observation, which I know is somewhat non-Stata-ish, and I am fully prepared
to be mocked for my inefficient coding.
The following works if you don't have any ties. That is, if, within any
farm, any of the enter times are the same as any of the exit times, then the
results might not be what you expect. You have to decide what you want to do
with ties. I.e., does a cow that arrives on the same day as another leaves
get counted as being in the pen for that day? With this code, an arriving
cow only gets counted on the next day. This would be true if the leaving
cows leave in the morning and the arriving cows arrive in the afternoon. The
code could accomodate the opposite by fiddling with the inequalities in line
10.
gen count=.
levels farmID
local levels = r(levels)
foreach farm of local levels {
gen marker = (farmID==`farm')
gsort -marker +cowID
count if marker==1
local N = r(N)
forvalues obs = 1/`N' {
count if enterPen<exitPen[`obs'] & exitPen>exitPen[`obs'] &
farmID==`farm'
replace count = r(N) in `obs'
}
drop marker
}
Lee
Lee Sieswerda, Epidemiologist
Thunder Bay District Health Unit
999 Balmoral Street
Thunder Bay, Ontario
Canada P7B 6E7
Tel: +1 (807) 625-5957
Fax: +1 (807) 623-2369
[email protected]
www.tbdhu.com
> -----Original Message-----
> From: Nick Cox [SMTP:[email protected]]
> Sent: Thursday, January 30, 2003 2:19 PM
> To: [email protected]
> Subject: RE: st: RE: counting across individuals within groups
>
> Buzz Burhans
> >
> > I still can't get what I need. What you suggested gives me
> > a count of all
> > the cows on a farm that have entered the pen cumulative as
> > of the date
> > they enter, and unconditioned on whether the other cows
> > that entered the
> > pen on a given farm are still present in the pen when a
> > particular cow
> > leaves. Note that farm and pen are different, all cows
> > with the same
> > farmid are present on the farm, but only a subset are
> > present in the pen on
> > any given date, and the cohort present on the day any given
> > cow exits the
> > pen is conditional on the dates the other cows entered and
> > exited. It is a
> > count of that subset cohort that I need, and presence in
> > the cohort may be
> > different for each "ExitiingCow" and is conditional on
> > several factors
> > relative to each unique "ExitingCow" as follows:
> >
> > 1. farmid is matched with farmid of "ExitingCow"
> > AND
> > 2. subset cohort cows must have enterPen dates < exitPen
> > dates of the
> > particular ExitingCow I am making the count for (i.e.
> > cohort cows entered
> > the pen before the ExitingCow left it)
> > AND
> > 3. subset cohort cows must have exitPen dates > exitPen date of the
> > ExitingCow (i.e. cohort cows had to still be present in the
> > pen when
> > ExitCow left.
> >
> > that is, cohort cows have to be on the same farm, enter the
> > pen before
> > ExitCow leaves it, and still be there when ExitCow leaves it.
> >
> > If it helps to make sense of what I am after, I am
> > interested in the
> > association of population density in a pen at the day of
> > exit and outcomes
> > for cows that are in certain management pens, i.e.
> > maternity or treatment
> > pens. I have dates each cow on a farm entered or left the
> > pens in question,
> > what I need is to figure out a count of the cohort present
> > at each cows
> > exit from the pen.
> >
> > . list farmid cowid enterPEN exitPEN enteredsofar
> > Present if farmid ==3,
> > table clean
> >
> > farmid cowid enterPEN exitPEN entere~r
> > Count(CowsinPenOnHerExitDate)
> > 50. 3 19 01/04/02 01/18/02 1 3
> > 51. 3 18 01/04/02 01/22/02 2 2
> > 52. 3 20 01/04/02 01/22/02 3 2
> > 54. 3 22 01/18/02 02/04/02 5 1
> > 55. 3 581 03/01/02 03/24/02 11 2
> > 56. 3 25 03/01/02 03/02/02 7
> > 57. 3 26 03/01/02 03/21/02 9
> > 58. 3 582 03/01/02 03/22/02 10
> > 59. 3 579 03/01/02 03/12/02 8
> > 60. 3 580 03/08/02 04/01/02 12
> > 61. 3 583 03/27/02 04/03/02 13
> > 62. 3 13 03/27/02 04/06/02 14
> >
> > The entere~r is the var that resulted from my test of your
> > code. The
> > "Count" is what I am after. For example, look at cowid
> > 581. 10 cows
> > entered before she exits, but 4 cows that entered before
> > she did exited
> > before she exits, so they are not in her cohort at exit. Of
> > the 6 other
> > cows that entered before she exited, 4 not only entered,
> > they also exited
> > before she did, so they are also not there when she exits.
> > Only cow 580
> > entered before she exited, but had not exited yet, so the
> > cohort at her
> > exit is 581 and 580.
> >
> > Thanks for following my ruminations.
>
> I can not hold all your detail in my head
> at the same time, but I think I can move closer
> to where you want to be.
>
> I suspect you wouldt benefit from mapping to
> a different data structure.
>
> First, -rename- -enterPEN- and -exitPEN-
>
> . rename enterPEN PEN1
> . rename exitPEN PEN2
>
> Now -reshape-
>
> . reshape long PEN , i(farmID cowID)
>
> You get something like
>
> +---------------------------------+
> | farmID cowID _j PEN |
> |---------------------------------|
> 1. | 3 13 1 Mar/27/02 |
> 2. | 3 13 2 Apr/6/02 |
> 3. | 3 18 1 Jan/4/02 |
> 4. | 3 18 2 Jan/22/02 |
> 5. | 3 19 1 Jan/4/02 |
> |---------------------------------|
> 6. | 3 19 2 Jan/18/02 |
> 7. | 3 20 1 Jan/4/02 |
> 8. | 3 20 2 Jan/22/02 |
> 9. | 3 22 1 Jan/18/02 |
> 10. | 3 22 2 Feb/4/02 |
> |---------------------------------|
> 11. | 3 25 1 Mar/1/02 |
> 12. | 3 25 2 Mar/2/02 |
> 13. | 3 26 1 Mar/1/02 |
> 14. | 3 26 2 Mar/21/02 |
> 15. | 3 579 1 Mar/1/02 |
> |---------------------------------|
> 16. | 3 579 2 Mar/12/02 |
> 17. | 3 580 1 Mar/8/02 |
> 18. | 3 580 2 Apr/1/02 |
> 19. | 3 581 1 Mar/1/02 |
> 20. | 3 581 2 Mar/24/02 |
> |---------------------------------|
> 21. | 3 582 1 Mar/1/02 |
> 22. | 3 582 2 Mar/22/02 |
> 23. | 3 583 1 Mar/27/02 |
> 24. | 3 583 2 Apr/3/02 |
> +---------------------------------+
>
> Now let us mark 1 for each entry and -1
> for each exit.
>
> . gen change = cond(_j == 1, 1, -1)
>
> If we cumulate that in day order we
> get a tally #beasts at the end of
> each day
>
> . bysort farmID (PEN) : gen no = sum(change)
>
> . bysort farmID PEN : replace no = no[_N]
>
> . l
>
> +-----------------------------------------------+
> | farmID cowID _j PEN change no |
> |-----------------------------------------------|
> 1. | 3 19 1 Jan/4/02 1 3 |
> 2. | 3 20 1 Jan/4/02 1 3 |
> 3. | 3 18 1 Jan/4/02 1 3 |
> 4. | 3 22 1 Jan/18/02 1 3 |
> 5. | 3 19 2 Jan/18/02 -1 3 |
> |-----------------------------------------------|
> 6. | 3 18 2 Jan/22/02 -1 1 |
> 7. | 3 20 2 Jan/22/02 -1 1 |
> 8. | 3 22 2 Feb/4/02 -1 0 |
> 9. | 3 581 1 Mar/1/02 1 5 |
> 10. | 3 26 1 Mar/1/02 1 5 |
> |-----------------------------------------------|
> 11. | 3 582 1 Mar/1/02 1 5 |
> 12. | 3 579 1 Mar/1/02 1 5 |
> 13. | 3 25 1 Mar/1/02 1 5 |
> 14. | 3 25 2 Mar/2/02 -1 4 |
> 15. | 3 580 1 Mar/8/02 1 5 |
> |-----------------------------------------------|
> 16. | 3 579 2 Mar/12/02 -1 4 |
> 17. | 3 26 2 Mar/21/02 -1 3 |
> 18. | 3 582 2 Mar/22/02 -1 2 |
> 19. | 3 581 2 Mar/24/02 -1 1 |
> 20. | 3 13 1 Mar/27/02 1 3 |
> |-----------------------------------------------|
> 21. | 3 583 1 Mar/27/02 1 3 |
> 22. | 3 580 2 Apr/1/02 -1 2 |
> 23. | 3 583 2 Apr/3/02 -1 1 |
> 24. | 3 13 2 Apr/6/02 -1 0 |
> +-----------------------------------------------+
>
> For each cow, you want to know how many were in
> the pen when she left
>
> . bysort farmID cowID (_j) : replace no = no[_N]
>
> Throw away -change- and you -reshape- back.
>
> . drop change
>
> . reshape wide
> (note: j = 1 2)
>
> Data long -> wide
> ----------------------------------------------------------------------
> -------
> Number of obs. 24 -> 12
> Number of variables 5 -> 5
> j variable (2 values) _j -> (dropped)
> xij variables:
> PEN -> PEN1 PEN2
> ----------------------------------------------------------------------
> -------
>
> . l
>
> +---------------------------------------------+
> | farmID cowID PEN1 PEN2 no |
> |---------------------------------------------|
> 1. | 3 13 Mar/27/02 Apr/6/02 0 |
> 2. | 3 18 Jan/4/02 Jan/22/02 1 |
> 3. | 3 19 Jan/4/02 Jan/18/02 3 |
> 4. | 3 20 Jan/4/02 Jan/22/02 1 |
> 5. | 3 22 Jan/18/02 Feb/4/02 0 |
> |---------------------------------------------|
> 6. | 3 25 Mar/1/02 Mar/2/02 4 |
> 7. | 3 26 Mar/1/02 Mar/21/02 3 |
> 8. | 3 579 Mar/1/02 Mar/12/02 4 |
> 9. | 3 580 Mar/8/02 Apr/1/02 2 |
> 10. | 3 581 Mar/1/02 Mar/24/02 1 |
> |---------------------------------------------|
> 11. | 3 582 Mar/1/02 Mar/22/02 2 |
> 12. | 3 583 Mar/27/02 Apr/3/02 1 |
> +---------------------------------------------+
>
> Some details remain ambiguous. Are you counting the cows
> _on each day_ before or after entries? before or after
> exits?
>
> My presumption remains that the tool of choice
> is -by:- but I have added the presumption that these
> matters are better done with a long data structure.
>
>
> Nick
> [email protected]
>
> *
> * For searches and help try:
> * http://www.stata.com/support/faqs/res/findit.html
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
