[Babase] Update trouble - Demog notes cannot have no note (no nulls allowed)?

Karl O. Pinc kop at meme.com
Tue Sep 7 14:13:03 EDT 2010


On 09/07/2010 12:46:46 PM, Niki Learn wrote:
> -----Original Message-----
> From: babase-bounces at eeblistserv.Princeton.EDU
> [mailto:babase-bounces at eeblistserv.Princeton.EDU] On Behalf Of Karl 
> O.
> Pinc
> Sent: Tuesday, September 07, 2010 1:02 PM
> To: The Baboon Database Project
> Subject: Re: [Babase] Update trouble - Demog notes cannot have no 
> note
> (no
> nulls allowed)?
> 
> On 09/07/2010 11:22:14 AM, Niki Learn wrote:
> > Yes, most of the M records in babase seem to be unnecessary and
> there
> > are at
> > least a small number that conflict with demography notes that are
> not
> > yet in
> > babase.  I expect I will delete most of these Ms when I do the 
> demog
> > note
> > backfill entries.  There do, however, seem to be a small number of
> > "real"
> > Manual entries where, for example, we don't have an actual census
> > datum but
> > we know that an animal was in another group prior to arriving in a
> > study
> > group and so we manually place the animal in that other group prior 
> > to
> > their
> > arrival in a censused group (such as when a previously unknown
> animal
> > is
> > seen coming from another group or an easily recognizable but 
> unnamed
> > animal
> > known to be in a peripheral group shows up in a study group).  The 
> M
> > then
> > provides a sense that the datum is a result of interpretation of
> > circumstantial evidence rather than being an actual census point.
> 
> You probably want an 'N' in these cases, because 'N' does not
> interpolate whereas 'M' does.
>
> Um...?  I've never used N.  There are only two baboons in babase with
> N
> records, all from the early '90s.  I have no idea why they have those
> (each
> of the two has a bunch of them).  Maybe when I am not in the midst of
> update
> I will look them up and see if there is any record of a reason.
>

The way I think about N is that you use it when you're not recording
absences.  With no absences interpolation does not really work
so you're stuck with telling the system every single day you
want an individual to be present.  This may or may not be
correct depending on what your expecting regarding absences.

No one has come up with
a good use-case, but N exists so that you can turn off interpolation
when you think it's not needed or will do the wrong thing.

> I don't see a problem with having interpolated points for the
> scenarios I
> mentioned above.

Ok.

>  In the case of an animal being seen joining from a
> nonstudy group, they would get the M record for the day before - 
> since
> this
> scenario typically happens in the morning when the groups have slept
> near
> each other, this is pretty straightforward and I don't think
> interpolating
> that manual entry is any different from interpolating any other 
> census
> point. 

Right.

> Ditto for an animal that is known to have been in a nonstudy
> group
> before arriving.

I guess that depends on what "known" means.

>  Occasionally (as recently happened with a couple
> animals
> that had been in Olkenya group), the team has actually included a
> descriptor
> of the known animal in the other group census data so you can go back
> and
> apply their new name to their record retroactively, but at other 
> times
> we
> just know that they were in the group and don't have any specific
> census
> records for them in that group (such as with the same males when they
> were
> in Stud's prior to joining Olkenya).  So what I have been thinking
> works
> best is to give them a manual census record on the last day they were
> absent
> from the study group, since that is the last date on which we know
> they were
> not there - again interpolation between last absence and first
> presence in
> the study group seems perfectly reasonable.

Sounds good to me.

> Btw, to make a manual entry I enter it in one of the normal
> fashions (census or demog) and then change it to manual, so I'm
> thinking
> this part of the definition - "when there is no other way to get the
> data
> in" - must be for babase 1.0 and no longer relevant?

That's not how I expected manual entries to get in, although you
can do it that way.  I figured manual entries would be entered
manually.  Using an SQL INSERT statement or adding a row with
PPA.

> 
> I'm more puzzled/concerned by how to get babase to interpolate
> correctly
> when animals are seen in two groups on the same day.  I have to pick
> one
> group for them to be in (and there doesn't seem to be any rule about
> which
> to pick) but then they get interpolated on either side of that even
> though
> we know they were in the other group the day before (at least in some
> cases,
> like if they slept with that group and then came over to the other
> group).

We just decided by fiat that an individual can only be in one group on
any given day.  

> This is another place we should be using M but where I think it has
> only
> been used sporadically.

Yes.  If you really know that he was in the group on the prior day.

>  Perhaps the rule would be to put them in the
> second
> group they were seen in but give them an M record in the first group
> the day
> before?

Sure.  Sounds sane.

>  Also, I find it annoying that nonstudy group animals keep
> reverting
> to group 9 in between censuses when the censuses are too far apart 
> (as
> they
> often are).  I guess they could be moving around during the time 
> we're
> not
> seeing them, but they could be doing that during the time they do get
> interpolated too...

Sure.  We made some arbitrary decisions as to how interpolation would
work.

>  So it seems like it would be nice for them to
> stay in
> the nonstudy group between censuses if they are present at both ends. 

The decision was to treat non-study groups the same as study groups.



> But I
> guess it is too complicated to have different rules for animals in
> different
> groups (with some groups changing category over time).

Yeah.  The general feeling is that interpolation is complicated
enough as it is.  The N Status is there so you can make exceptions.



Karl <kop at meme.com>
Free Software:  "You don't pay back, you pay forward."
                 -- Robert A. Heinlein




More information about the Babase mailing list