[Babase] Re: Enhancements to Ranker

Lacey Maryott lacey.maryott at duke.edu
Tue Dec 4 10:19:56 EST 2007



kfenn wrote:
> Michael Altmann wrote:
>> Susan and Karl both pointed me towards the mailing list, so perhaps 
>> in the
>> interest of efficiency some one person could flesh out, and post, the
>> details on the following TODOs
>>
>>
>> High Priority
>> Fix Anomalous interaction occurances (requires some extensive code 
>> clean up)
>>
>> Q: More detail?
We found a lot of interactions showing up in the new ranker matrices 
that weren't on the old ranker matrices, and that didn't exist in the 
database. They were just 1 interaction usually and they existed right 
around the diagonal line. They also only occurred between juvenile 
males. It is the case that sometimes they were symmetrical, 1 win for 
each individual, so it wouldn't have changed the decision we made, but 
that isn't true for every case, and in every case, the actual 
interactions could not be found in the database. Some specific examples 
can be found in an e-mail I sent in the babase archive on 12 Oct 2007 
Subject: Ranker Problem.
>> Fix Rnktypes queries so that the correct subset of individuals is 
>> produced
>> for each rnktype. Q: Where would I get the correct queries?  Which 
>> ones are currently wrong?
>>   
> Lacey ran some tests on these two issues above.  I think the problems 
> were primarily male.  She identified (and possibly cleaned up?) the 
> code for the second issue, I think.  Lacey, can you fill in the 
> details here?
The query of concern here was the ADM rnktype, and I did fix this query 
in babase, so as far as I know, that one is taken care of.
>
>> Coding flexibility into the time intervals for which the matrices are
>> produced. Q: Does this mean the user picks a start and stop date?
>>   
> Yes, we want the flexibility to pick any stop and start dates we choose.
>> Ranks based upon previous month's ranks, not in alphabetical order. 
>> This was
>> not discussed during the CC, but is important   
> You don't have a question here...is this item clear?  At a minimum, we 
> need the program to generate each new matrix using the existing 
> rankings of the most recent month in the database.  My WISH is to be 
> able to specify which month/year rank data I want the program to use 
> to create the matrix.
>> Lower Priority
>> Alternatively place the stats in a seperate, recallable window. A 
>> 'Fit' vs
>> '100%' zoom option might work too, but these options are less urgent.
>>   
> That's for Duke to explain, but we were having problems not seeing all 
> the snames in the columns and rows.
Yes, on this, the problem is when the groups get to large, and there are 
a lot of names across the top of the matrix, the names on the left side 
get squished to a first letter and ''...'' which, in groups where 
animals are often named with the same first letter as their mother and 
siblings, is too vague.  We were hoping that, if we wanted we could 
switch back and forth between having a scroll bar with a 100% view, such 
that no names are getting smushed, or chopped, we just have to scroll 
over to see the whole matrix, or a 'fit' option like powerpoint has, so 
that we can see the whole matrix, but slightly smaller.
>> Automated extraction of rank reversals Q: More detail?
>>   
> I can't send attachments on this email list, but I will directly email 
> you a sample matrix with the rank reversals circled and a brief 
> explanation of what I hope to get.
>> Display Month/Year of Currently Displayed Ranking
>> Q: Any particular place you had in mind?
>>   
> Placement isn't so important to me but I want the parameters defining 
> the matrix to always appear somewhere when the matrix is viewed.
> 1)  In the on-screen environment, I don't care if the month/year are 
> in a separate popup window with the stats. Presumably, I will generate 
> the matrix and fool with it during one sitting and remember the 
> month/year, or can easily re-open the popup window if I need to be 
> reminded.  I don't know how Lacey and Susan feel about this.
> 2)  In any electronic output file we need the month year and all 
> associated stats to be saved within the same file.  The original 
> ranker allowed you to save the output to a text file. I'm not aware of 
> a way to save the matrix and stats output right now, but it would be 
> essential to have all month year and stats stored with the matrix  
> (i.e. I don't want to ever open a soft copy of a matrix but then not 
> know what parameters I used to create it)
> 3)  In any printed hard copy we need the month year to show up, along 
> with the associated statistics.  It would be on one page, if possible 
> (when I send the output with the rank reversals, you can see how it 
> works now.)
>
I have no real opinion on it's placement, what Tabby said is good.
> I'll let Lacey and Susan put their 2 cents in.  We can still do a 
> conference call once you read this and have a chance to look at the 
> attachments I send.
>
>
> Tabby
>
>

-- 
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