Re: st: Number of people present by date and time

[Nick Cox <njcoxstata@gmail.com>]

Should be easier than I implied. Even if a unique identifier doesn't
exist for each observation, you just create one. For a _big_ dataset,
be careful on variable type.

I am assuming that -arrival- and -depart- are Stata date-times.

gen long obsid = _n
expand 2
bysort obsid : gen inout = cond(_n == 1, 1, -1)
by obsid : gen double time = cond(_n == 1, arrival, depart)
sort time
gen present = sum(inout)

Two simple checks on logic and data quality

1. The number in the clinic should never be negative.

2. The number in the clinic should be zero when the clinic is closed.

Nick

On Thu, Nov 29, 2012 at 2:01 PM, Nick Cox <njcoxstata@gmail.com> wrote:
> Each observation is, I gather, a patient. One technique is to make
> each observation an arrival or departure. For a very simple toy
> dataset with just times for one day:
>
> . l
>
>      +-----------------------+
>      | arrival   depart   id |
>      |-----------------------|
>   1. |    1000     1100    1 |
>   2. |    1030     1200    2 |
>   3. |    1230     1300    3 |
>      +-----------------------+
>
> . expand 2
> (3 observations created)
>
> . bysort id : gen inout = cond(_n == 1, 1, -1)
>
> . by id : gen time = cond(_n == 1, arrival, depart)
>
> . sort time
>
> . l
>
>      +--------------------------------------+
>      | arrival   depart   id   inout   time |
>      |--------------------------------------|
>   1. |    1000     1100    1       1   1000 |
>   2. |    1030     1200    2       1   1030 |
>   3. |    1000     1100    1      -1   1100 |
>   4. |    1030     1200    2      -1   1200 |
>   5. |    1230     1300    3       1   1230 |
>      |--------------------------------------|
>   6. |    1230     1300    3      -1   1300 |
>      +--------------------------------------+
>
> . gen present = sum(inout)
>
> . l, sep(0)
>
>      +------------------------------------------------+
>      | arrival   depart   id   inout   time   present |
>      |------------------------------------------------|
>   1. |    1000     1100    1       1   1000         1 |
>   2. |    1030     1200    2       1   1030         2 |
>   3. |    1000     1100    1      -1   1100         1 |
>   4. |    1030     1200    2      -1   1200         0 |
>   5. |    1230     1300    3       1   1230         1 |
>   6. |    1230     1300    3      -1   1300         0 |
>      +------------------------------------------------+
>
> This is only one trick, and others will depend on your data. For
> example, if your clinic is only open daily, you may be able to, or
> need to, exploit that. If patients can come to a clinic more than once
> a day  that will provide a complication.
>
> All told, you should not need loops here. The two keys are likely to
> be (1) the best data structure (2) heavy use of -by:-.
>
> Nick
>
> On Thu, Nov 29, 2012 at 1:35 PM, Simon <scmoore.lists@googlemail.com> wrote:

>> This is quite possible a rather naive question, but for some reason I am
>> stuck.
>>
>> I have data from a clinic. I have the time each patient checks in
>> (arrdatetime), the time they leave (depdatetime) and the time taken to
>> first consultation (waittime) in minutes. What I would like to do is
>> compare the number of people in the clinic for each patient at
>> arrdatetime with waittime.
>>
>> So far the best I can come up with is to write a loop, going through
>> every patients' arrdatetime and counting up those whose arrival and
>> departure times span this value. But I have rather a lot of data and
>> this seems terribly inefficient.
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/