[Babase] Re: Enhancements to Ranker
Lacey Maryott
lacey.maryott at duke.edu
Tue Jan 8 10:36:24 EST 2008
Jun Yang wrote:
> On Dec 20, 2007 2:05 PM, Lacey Maryott <lacey.maryott at duke.edu> wrote:
>
>
>> First, I created a matrix with rank dates 12-01-2005 to 12-31-2005
>> as the basis and with 01-01-2006 to 01-31-2006 as the interaction matrix
>> dates for ALM ranks for grp 1.10. The actual ranks for this grp for Dec
>> 05 do not match the initial output in ranker. I made a table here
>> <https://papio.biology.duke.edu/babasewiki/RankerDiffs#preview> on the
>> wiki so that you can see the difference. The order produced by Ranker
>> doesn't seem to resemble at all the ranks for Dec 05.
>> Also, the same ranking produced the list of individuals who had been
>> present in December, not those present in January. NET for example, was
>> in 1.10 in Dec 2005, but was not as of Jan 2006, and thus, should not
>> show up in the matrix, (Right?).
>> Lastly, I really like the new header at the top of the window which
>> displays the group and date of the matrix, but I think it is important
>> to distinguish that while the ranking is based on DEC 05 Ranks, that
>> interactions are from Jan 06. Right now the header for the interactions
>> taking place in Jan 06 Reads "Nyayo's ALM on 2005-12-01".
>>
>
> I think these issues are all rooted in our interpretation of "dates of ranking."
>
> 1. The "ranking start/end dates" refer to the month that the ranking is for
> (not the month with an existing ranking from which we derive the
> current ranking).
> For example, if the ranking start/end dates specify Dec 05, then we
> look for all members
> of Nyayo's ALM during Dec 05. These members, and only these members, will
> show up as headers of the matrix (regardless of whether they participate in any
> interactions during the "interaction matrix start/end dates"--see below).
>
> When we save the ranking to babase, the current ordering of members (matrix
> headers) will be recorded in the babase Ranks table, with rnkdate set
> to the ranking
> start date. (The contents of the interaction matrix is never written to babase).
>
> 2. The "interaction matrix start/end dates" refer to the time period during
> which we count the number of interactions among the members selected
> for ranking according to the "ranking start/end dates" (see above).
> For flexibility,
> this time period can be different from the ranking start/end dates.
>
Ok, so to clarify, If I want to rank all members of grp 1.1 for January
'06 based on interactions which took place in Jan 06, both of the date
fields should be for Jan 1 - Jan 31 06?
> 3. When you create a ranking (using "Create/load ranking from Babase"
> on the welcome
> menu), the members will always listed in alphabetical order. The ranker doesn't
> check whether a ranking for the specified ranking start/end dates already exists
> in the babase Ranks table---if there is one, it is almost certainly
> different from
> the alphabetical order. This is what you observed at
> <https://papio.biology.duke.edu/babasewiki/RankerDiffs#preview>.
>
> 4. However, after you've created a ranking, you can always select
> "Load/incorporate
> ranking from Babase" from the menu bar above the interaction matrix. The ranker
> will try to locate the an existing ranking in the babase Ranks table
> for the specified
> ranking start/end dates (and if none exists, the latest one before the
> ranking start
> state). The ranker will then attempt to reorder your current ranking
> according to
> the existing ranking, and ask you for confirmation. This is how you
> load an existing
> ranking. In the example you gave, if you select "load/incorporate,"
> the ranker will
> prompt you to change the current ranking to the existing one.
>
> Perhaps a better interface would be to directly load the existing
> ranking (if any) when
> "Create/load ranking from Babase" (in 3), but currently you can
> accomplished that by an
> extra, explicit step (4).
>
I think it is definitely the case that the old ranker loaded them in the
previous months rank order. As I have never needed to actually create
matrices, Susan, Tabby, do either of you have any input on whether it
would be better to have it automatically load previous rankings, or to
load alphabetically and then have the explicit step to ask it to order
the individuals according to the previous months ranking? From what I
know of the process, I don't forsee there ever being a time when we
won't want the matrices ordered according to the previous month's
ranking, but like I said, I haven't run enough of these myself to know
whether those circumstances would present themselves that we would want
them listed alphabetically, or how much it would slow down the process
to have to reorder them each time.
> Back to your example of NET. I think NET should shown as a column header even
> though he didn't participate in any interaction, because the order of
> column headers
> is what determines the current ranking. If we omit NET from the
> headers, we wouldn't
> have any way to specify how NET is ranked relative to the others.
>
This is totally true. My confusion was with the two different date
types, which created the issue of NET vs no NET. In the case that an
individual is in the group but doesn't interact (which is extremely
common) we definitely want his rank. We just don't want him in the
ranking if he isn't in the group at all.
> As for the the window title of the main interface showing the
> interaction matrix, I
> can update it to show both ranking month and interaction matrix start/end dates
> to avoid confusion.
>
Sounds good, although I think my confusion on this is cleared up at
least :-P Just need to pick the same date ranges for both fields in
most cases, right?
> --- Jun Y.
> _______________________________________________
> Babase mailing list
> Babase at www.eco.princeton.edu
> http://www.eco.princeton.edu/mailman/listinfo/babase
>
>
--
Lacey Maryott
Alberts Lab
Department of Biology
Duke University
ph: 919-660-7306
fax: 919-660-7293
Lacey.Maryott at duke.edu
More information about the Babase
mailing list