Re: st: Duration data- count number of spells with ref to current spell

[Nick Cox <njcoxstata@gmail.com>]

The cumulative or running sum is given by -sum()-. For a toy dataset
apply this separately.

. bysort id : gen cumact = sum(activity)

. l

     +------------------------+
     | id   activity   cumact |
     |------------------------|
  1. |  1          1        1 |
  2. |  1          3        4 |
  3. |  1          5        9 |
  4. |  1          7       16 |
  5. |  1          9       25 |
     |------------------------|
  6. |  2          2        2 |
  7. |  2          4        6 |
  8. |  2          6       12 |
  9. |  2          8       20 |
     +------------------------+

The activity before any event is

by id : gen before = cumact[_n-1]

The activity after any event is

by id : gen after = cumact[_N] - cumact

In your case, you will want to sort by time, but the main idea should
carry over.

Nick
njcoxstata@gmail.com


On 26 August 2013 20:51, Sriram Narayanamoorthy <nsriram@utexas.edu> wrote:
> Hello everyone,
>
> This question is on data generation when working with a
> discrete-duration datasets. The dataset I have includes all the
> activities undertaken by individuals in a 24 hour period. The day is
> divided into 48 half-hour intervals. The start interval of a spell is
> denoted by a variable called eventStart. The activity type is denoted
> by the variable eventType.
>
> The task of counting the number of activities of a particular type
> undertaken in a day by the person is quite straightforward using egen.
>
> However, I want to generate a variable as the number of activities of
> a particular type undertaken by the individual before/after the
> current activity. How do I add the "Before" and "After" dimension to
> my variable? To be sure, I want to compare the eventStart of my
> current record with all other records for that person and count the
> number of records of a particular eventType only if eventStart is less
> than (for Before) or greater than (for After) the current start
> period.
>
> Any thoughts? Is there a mata approach to this? Thanks much for the inputs!
>
> Thanks,
> -Sriram
> *
> *   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/
*
*   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/