[Babase] MEMBERS Comparison - Final Summary
Catherine Markham
babase@www.eco.princeton.edu
Fri, 30 Sep 2005 13:58:37 -0400
Hi Karl,
I finished with the MEMBERS comparison just a little while ago. Below
are a couple questions that came up that I can't quite explain.
Everything else - by far the majority of discrepancies - made sense
(yippee!) or I've already emailed you about them.
Thanks,
Catherine
Questions:
1. Several individuals had only rows in the new MEMBERS table despite
having no rows in CENSUS. These individuals include: A21, B11, CS1*,
D13*, DM1*, MN1*, MT5*, NZ5*, R14, RH9, and UT4*.
All of the ones with asterisks were on my list of conversion errors from
Leah (corrections have been discussed, but I still need to make the
changes). I'm not sure what is up with the others, though - I'll have
to look into them more carefully. What the program is doing in all
cases is giving a single row for the baboon's birth in MEMBERS, despite
there being no data in CENSUS. The only exception is A21 because this
is the only individual in the list who also has a death date - in this
case, there are rows from birth through death in MEMBERS, again without
any info in CENSUS. All of this seems logical, even if it is a bit
weird to have MEMBERS entries when nothing is listed in CENSUS. Karl,
am I working through this one right?
2. Lots of "glitches" arising from the partially loaded Wn0303b file
showed up as rows in MEMBERS not matching CENSUS. I flagged these to go
back and recheck after that census file for Weaver's on 5 March 2003 can
be uploaded completely.
3. The last few records in the comparison txt file you sent me start
out for ZUM - the last line itself is nothing more than a plus sign. Is
this the correct end? No real errors to look at for ZUM?
Common discrepancies:
1. The most common discrepancies had to do with interpolation around a
baboon's death due to changes in the interpolations rules.
2. Another common discrepancy had to do with the fact that no days
placing a baboon in Group 9 were reflected in the comparison. These
were not true differences in the tables, but required checking the new
MEMBERS to be certain the rows did in fact exist.
3. The mistake with many of TIT's rows in CENSUS where status had been
"L" and is now changed to "N" (to match TIT's rows in MEMBERS).
How I did it (just for archive):
For each baboon with an error (whether it involved a plus or minus or
both), I looked up the baboon's sname in BIOGRAPH and checked to see if
the dates around the discrepancy had to do with birth, death, or
neither. Most of the problems around the birth date dealt with animals
born into Group 9 (their first 14 days post-birth showed up with a minus
sign since Group 9 rows in the new table were ignored in the
comparison). For these, I checked the new MEMBERS table to make sure
that the expected rows were indeed present.
For the deaths, I checked the old CENSUS table for the last confirmed
presence and first confirmed absence (and flagged records where first
confirmed absence was not entered - still need to correct these). I
then calculated what the interpolation should be by hand and checked the
new MEMBERS table to make sure it matched correctly.
For neither, these were mostly just cases where I needed to verify that
placement the animal was actually in Group 9 in the new MEMBERS table.
There were just a couple cases where running reinterp on the old MEMBERS
table sorted out any discrepancies.