[Babase] Important update to Ranker Tweaking
kfenn
kfenn at princeton.edu
Mon Jul 21 11:42:16 EDT 2008
Jun Yang wrote:
> Hi Tabby,
>
> I just double-checked the rnktypes tables in babase_test and babase,
> and made sure that they are consistent.
>
> Can you elaborate on what exactly is unexpected? Do babase_test and
> babase have identical contents for the ranks table (and other tables)?
> If they are, then the behavior of the ranker should be consistent.
>
> When you said below that the ranker creates an order that does not
> match the one you saw in babase_test. What exactly do you mean?
>
>
Hi Jun,
The tables are identical except in some cases when we haven't copied
babase to babase_test for some reason. It was the first thing I looked
at because I know the tables don't always match. That doesn't seem to
be the source of the present problem.
If I run the query
select * from ranks where grp = 1.21 and rnktype = 'ALF' and rnkdate >
'2006-11-30' and rnkdate < '2007-01-01' order by rank
I get the same output in phppgadmin whether I do the selection in babase
or babase_test. In other words, the rank table for Dec06 seems to be
the same in both databases.
So let's just look at the first four individuals in the rank order for
Dec 06 (in both babase and babase_test):
DUI 1
ABB 2
ACA 3
ECH 4
I used babase as the database for Ranker. I ran a matrix for Jan 07,
but specified that I want to use the Dec 06 rank order. I got a matrix
in which the first four individuals followed the rank order above, as
expected.
However, when I use babase_test, either allowing it to use Dec 06 via
the default OR even manually going into the Rank menu and choosing Dec 1
2006 from the calendar, I get the following rank order for the first 4
individuals:
ACA 1
DUI 2
ABB 3
ECH 4
Try it for yourself and let me know if you get different results.
One further thing I noticed is that the dialogue box "Incorporate
previous ranking" always says:
Ranking for ALF: Omo's on 2006-12-01 from Babase
ACA DUI ABB ECH EAG OXY OPH ONY EUD OPA ORI OCT OFR OAS
Is Ranker actually reading from babase_test when I specify, but the
text is just referring to babase_test generically as "babase"? Could
you change the text so that it says it is reading from babase_test?
Once you get past that intro screen, its difficult to tell which
database you are using.
> 1. Was there already an order stored for 2007-01-01 for that
> group/rank type in babase_test, but not in babase (or vice versa)? If
> that's the case, it's not the ranker's problem, because somebody has
> manually saved an order in one database but not the other, and that
> order may or may not be consist with the default (we expect users to
> make manual tweaks to orders).
>
> 2. If the above is not case, and you are saying that for both babase
> and babase_test, the ranker incorporated the same previous ranking
> (2006-12-01) but came with different default rankings, then one
> possible explanation is that the two databases are not completely
> synchronized, and certain factors affecting the default ranking differ
> across these two databases.
>
>
Yes, I am saying #2. I can't come up with a rational explanation for
why it is behaving this way since the ranks order in both database
tables apparently match.
Thanks,
Tabby
> --- Jun Y.
>
>
> On Tue, Jul 15, 2008 at 8:27 AM, kfenn <kfenn at princeton.edu> wrote:
>
>> Jun Yang wrote:
>>
>>> When you first connect to babase through ranker (in its welcome
>>> screen), you can select babase-test and then all reads and writes will
>>> go to babase-test instead of babase. Alternatively, you can also save
>>> the ranking to a file (on your computer), load it, and then then later
>>> save to a database.
>>>
>>>
>> Sorry, I'm the one who told Lacey it wasn't working. I had forgotten the
>> exact nature of my problem with this so I checked again this morning. Yes,
>> you can select which dbase to work with but connecting to babase_test
>> creates mistakes in the rank order. For instance, when I create and ALF
>> matrix for Omo in 2007-01-01, it asks if I want to use the most recent
>> 2006-12-01 ranking. If I accept this default, it creates an order that does
>> not match the one I see in babase_test (ACA is first but DUI should be
>> rank1).
>>
>> Tabby
>>
>>> --- Jun Y.
>>> _______________________________________________
>>> Babase mailing list
>>> Babase at www.eco.princeton.edu
>>> http://www.eco.princeton.edu/mailman/listinfo/babase
>>>
>>>
>> --
>> Tabby Fenn
>> Research Assistant
>>
>> Dept of Ecology and Evolutionary Biology
>> 401 Guyot Hall
>> Princeton University
>> Princeton, NJ 08544
>>
>> 609 258-6898 (Ph)
>> 609 258-2712 (Fx)
>>
>> _______________________________________________
>> Babase mailing list
>> Babase at www.eco.princeton.edu
>> http://www.eco.princeton.edu/mailman/listinfo/babase
>>
>>
> _______________________________________________
> Babase mailing list
> Babase at www.eco.princeton.edu
> http://www.eco.princeton.edu/mailman/listinfo/babase
>
--
Tabby Fenn
Research Assistant
Dept of Ecology and Evolutionary Biology
401 Guyot Hall
Princeton University
Princeton, NJ 08544
609 258-6898 (Ph)
609 258-2712 (Fx)
More information about the Babase
mailing list