[Babase] Question about statdates and conversion errors
Karl O. Pinc
babase@www.eco.princeton.edu
Wed, 31 Aug 2005 20:35:09 +0000
On 08/31/2005 02:36:45 PM, Leah Gerber wrote:
> Hello Susan, Jeanne, Karl, and Catherine,
>
> While going through some of the conversion errors an issue about
> statdates has come up. Some older statdates have been calculated as
> the midpoint between the last day an individual was marked present in
> a census and the first day the individual was marked absent in the
> next census of the same group. The definition for statdate when an
> individual is alive and absent in last census (Figure 12 in the
> documentation on interpolation) states that the statdate should be
> the last day where the individual is censused as present. When fixing
> conversion errors should I correct these dates to reflect the current
> definition of statdate? If we decide to correct some this way should
> I also go through all individuals that may have these midpoint
> statdates and correct them as well?
The conversion process is:
1) Load biograph, marking everybody alive and statdate = birthdate
(so interpolation can run fast and not have to interpolate out
all the way to the statdate for each census row loaded.)
2) Load census. Interpolation creates members and makes statdate
= last censused present date.
3) Load demog. Interpolation does likewise.
4) Update biograph, setting statdate, status, and dcause. Database
complains at this point if any rules are violated.
So, at step 4 if the individual is alive you should see an error
from the database if statdate != last censused present. Dead
individuals on the other hand can have any statdate >= last censued
present. So, you will be _required_ to correct all living (status = 0)
individuals. The database won't check dead individuals so if there's
funny statdates with those you'll have to find those and make any
corrections you see fit.
Note that members is unchanged because (now) interpolation will
interpolate past the statdate of living individuals halfway to
the next absence. (Which is probably why the statdates were
done the way they were.)
Karl <kop@meme.com>
Free Software: "You don't pay back, you pay forward."
-- Robert A. Heinlein