[Babase] REPSTATS design problems, a review
Karl O. Pinc
babase@www.eco.princeton.edu
Fri, 10 Sep 2004 14:37:21 -0500
Hi,
I went over this with Jeanne on the phone the other day,
but it seems good to write for the record, and because
we'll need a decision, even if that's to do nothing.
REPSTATS records reproductive state by individual by day.
In an effort to make it have records for as many days
as possible we decided it was ok for CYCLES to have
missing data, REPSTATS would continue to have records
for the days of those cycles. However, this has
some problems.
1) The first cycle starts with a T date. When there is
no T date in the first cycle, where should REPSTATS
start recording the individual as "cycling"? The
D date? What if there's no D date? If we use the
D date when there's no T date, it's odd that
the first cycles of different individuals would
start at different points in thier actual sexual cycle.
We have no requirement the first cycle have a T date.
Presently, when there's no initial T date, we don't
get a 'cycling' state in REPSTATS until we do get
some T date. The previous state (i.e no REPSTATS)
is continued until there's data.
2) This is not really a problem according to Steph and
the way the actual data exists, but we have the same
problem as 1 with post-luteal T dates. The luteal
state could go on until we happen to get a T date.
3) Cycles go from T date to T date, as they start with the
first T date, and first T date post pregnancy.
REPSTATS rows 'within' a "cycling" cycle have
a column for the CYCLES.Cid of the starting T date and
the CYCLES.Cid of the ending T date. Because we're
allowing T dates to be missing in the CYCLES, the cycles
with no T date will not ever be referenced in REPSTATS.
REPSTATS sees one long cycle. Without a T date, REPSTATS
does not know where to start/stop the cycle.
3a) When a conception cycle has no T date REPSTATS does not record
the conception CYCLES row as the cycle which starts the 'cycling'
state which preceeds the pregnancy. It's some earlier cycle.
4) The last 'state' of an individual is continued to the individual's
statdate, so after menopause the individual remains in the 'cycling'
state. This is not a new problem.
5) As a result of all of the above, the REPSTATS days-into-cycle
columns are pretty useless. Although I suppose they do provide a
clue when they are useless, when the sum of days-into-cycle and
days-remaining-in-cycle is absurd, you know the data is absurd.
Karl <kop@meme.com>
Free Software: "You don't pay back, you pay forward."
-- Robert A. Heinlein