Re: st: Changing spells to person-years

[David Kantor <kantor.d@att.net>]

At 01:30 PM 11/20/2006, Justin wrote:
> Hello trusty listservers,
>
>
> I have spell that that I want to convert into a long file. Using a 
> life history calendar, I have below four people who show changing 
> residences (rural/urban) across their life course. Each row 
> represents one person. For instance, person 1 lived in a rural area 
> from 0-3, an urban area from 4-6, rural from 7-15, and urban 16-20. 
> That person at the time of the survey was 20 years old. Person 3 
> lived in a rural area her whole life.
>
>
> cw2i1   cw2i1s cw2i1e cw2i2  cw2i2s cw2i2e cw2i3  cw2i3s cw2i3e 
> cw2i4  cw2i4s cw2i4e
>
> rural          0         3      urban         4         6      rural 
>        7         15    urban      16        20
>
> urban        0        30      rural        31       32       . 
>        .          .           .            .         .
>
> rural         0         38        .              .          . 
>  .              .           .         .           .          .
>
> urban        0        17      rural       18        54      urban 
>   55        57       .            .          .
>
>
> While I have shown how it looks for residence location, the same 
> exists for education and other variables. I know I could generate 
> hundreds and hundreds of new variables so that each cell in the 
> calendar represents a variable in STATA, but from reading the 
> listserv, it seems that there is an easier way, although it is not 
> clear to me. How can I turn this into a person-year file?
This is somewhat straightforward for this one attribute.  It's a 
-stack- or -reshape long- matter.

BUT, what happens when you combine other attributes.  Do they all fit 
into the same patterns of years? Or does each attribute have its own 
pattern of years?

If the latter, then you may need to first treat each attribute 
separately.  Then put them together with a -merge-.

Another matter: Is it always true that the end of a spell is 
start_of_next_spell -1 ?  (No gaps, no overlaps.)  If so, then you 
can eliminate the end dates. (Though, you may need to insert a dummy 
spell for each person to signify the end of the latest-known spell.) 
And then, later, in the -merge- step, the match-key variables will be 
person-id and start_year.  (But in any case, you should not include 
the end_dates in the merge.)

Under this scheme, the result will be that for each person, different 
attributes will all share the same pattern of spells, using a "fine" 
granulation of years -- where the start dates are the union of all 
start dates of the spells of the different attributes. Then, some 
attributes will not change at every spell transition.  A spell 
transition will occur when at least one attribute changes.

If you spell out the structure of your data in more detail, then we 
can better help you.
HTH
--David

*
*   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/