LISTSERV at the University of Georgia
Menubar Imagemap
Home Browse Manage Request Manuals Register
Previous messageNext messagePrevious in topicNext in topicPrevious by same authorNext by same authorPrevious page (December 2006, week 3)Back to main SAS-L pageJoin or leave SAS-L (or change settings)ReplyPost a new messageSearchProportional fontNon-proportional font
Date:         Mon, 18 Dec 2006 22:03:15 -0800
Reply-To:     David L Cassell <davidlcassell@MSN.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         David L Cassell <davidlcassell@MSN.COM>
Subject:      Re: Problem Ranking students in to deciles
Comments: To: gopilth@yahoo.com
In-Reply-To:  <1166406479.551544.220860@73g2000cwn.googlegroups.com>
Content-Type: text/plain; format=flowed

gopilth@yahoo.com wrote to me personally instead of to the list: > >David L Cassell wrote: > > gopilth@YAHOO.COM wrote back: > > > > > >David L Cassell wrote: > > > > datanull@GMAIL.COM sagely replied: > > > > > > > > > >On 12/11/06, gopilth@yahoo.com <gopilth@yahoo.com> wrote: > > > > >>Hi, > > > > >>I have problem ranking students into deciles (1-10) based on their > > > > >>score.Following is my sample data set > > > > >> > > > > >>id score > > > > >>1 512 > > > > >>2 512 > > > > >>3 618 > > > > >>4 620 > > > > >>5 620 > > > > >>6 625 > > > > >>7 650 > > > > >>8 651 > > > > >>9 700 > > > > >>10 701 > > > > >> > > > > >>I would appriciate any kind of help. > > > > >>Thanks in advance. > > > > >>Gopi > > > > >> > > > > > > > > > > > > > >Deciles 10 groups? > > > > > > > > > >data work.score; > > > > > input id score @@; > > > > > cards; > > > > >1 512 2 512 3 618 4 620 5 620 > > > > >6 625 7 650 8 651 9 700 10 701 > > > > >;;;; > > > > > run; > > > > >proc rank groups=10; /* numbered from 0 to 9 */ > > > > > var score; > > > > > ranks decile; > > > > > run; > > > > >proc print; > > > > > run; > > > > > > > > > > > > > In addition to Data _Null_'s sound advice, let me point out that > > > > ties have to be considered in cases like this. > > > > > > > > How are you supposed to handle things in cases like your bottom > > > > two deciles? > > > > What are your real data like? > > > > And why do you need deciles? > > > > > > > > HTH, > > > > David > > > > -- > > > > David L. Cassell > > > > mathematical statistician > > > > Design Pathways > > > > 3115 NW Norwood Pl. > > > > Corvallis OR 97330 > > > > > > > > > > Hi David, > > >I need deciles from score to give points to each student based on their > > >decile. there are ties in my data. will that be a problem? following is > > >sample of my real data. > > >id score > > >0829264 032 > > >0003392 034 > > >1228873 032 > > >1228907 032 > > >1230200 028 > > >1230242 032 > > >1230671 032 > > >1231406 039 > > >5108113 034 > > >8050171 036 > > >8074122 037 > > >8085227 030 > > >8152001 035 > > >8152977 037 > > >Please advise > > >Thank you, > > >Gopi. > > > > Why score students based on deciles at all? I don't follow that. > > I never did that when I was teaching. Are you sure that's a good > > approach? > > > > And what do you plan to do when you have ties? You have > > 5 of your 14 values above with the same value: 032 . You > > have to assign them all to the same decile. Do you want to > > screw them out of a decent grade and assign them the lowest > > decile possible? Or give them a big jump and assign them the > > highest decile possible? Or assign them an intermediate decile? > > you have to make decisions on this, and then PROC RANK can > > impose those decisions. But it would be totally unfair to give > > students with the same grade completely different deciles. > > > > As I look at your data, this just makes no sense. If you have > > values which are only a digit or two apart, but you end up > > having to assign them to very different deciles, then are you not > > cheating the students out of fair grades? > > > > HTH, > > David > > -- > > David L. Cassell > > mathematical statistician > > Design Pathways > > 3115 NW Norwood Pl. > > Corvallis OR 97330 > >

>Hi David, >Thanks for your response.It would be a great help if you could help me >resolve this problem. >My data has ties and as you said it would not be fair to give deciles >with ties, as I might end up giving different deciles to students with >same score. I do not want to manipulate the ranking by imposing proc >rank to put data with ties in a particular >group(highest/lowest/intermediate). I want all the students to have the >decile per their score, so I was thinking if following would be a good >approach for this problem. >1. Sort my data, so as to get a new data set with no ties and then, >2. Use proc rank to assign deciles. >3. Once I get the output, write a logic for whole data with ties as to >assign a decile to a score range (ex: if score ge 15 and lt 20 then >decile eq 5). > >Can you please advise the right way to get correct deciles to data with >ties. > >Thank you, >Gopi.

[1] Sorting is a huge waste of time. PROC RANK does that for you, internally.

[2] Killing records to prevent ties will make things hard to re-assemble afterward. Re-labeling students so they do nto appear to be tied will cheat the students of well-earned grades. That's even worse.

[3] PROC RANK handles ties properly. You just have to answer the original question I asked:

> > And what do you plan to do when you have ties? You have > > 5 of your 14 values above with the same value: 032 . You > > have to assign them all to the same decile. Do you want to > > screw them out of a decent grade and assign them the lowest > > decile possible? Or give them a big jump and assign them the > > highest decile possible? Or assign them an intermediate decile? > > you have to make decisions on this, and then PROC RANK can > > impose those decisions. But it would be totally unfair to give > > students with the same grade completely different deciles.

So. Do you want to assign tied grades to the average rank, the lowest rank, or the highest rank of the possible ranks that we would assign? This is like to make a big different in what grade these people get!

Either way, PROC RANK can do it for you.

[4] Why are you using deciles and ranks? This seems like a really bad way to assign grades. Can you justify this on the basis of distribution of student grades, or class performance measures, or anything? Right now, I have to conclude that you should NOT be doing this, and hence advising you on the ins and outs of PROC RANK is actually doing you (and those students) a disservice.

HTH, David -- David L. Cassell mathematical statistician Design Pathways 3115 NW Norwood Pl. Corvallis OR 97330

_________________________________________________________________ Your Hotmail address already works to sign into Windows Live Messenger! Get it now http://clk.atdmt.com/MSN/go/msnnkwme0020000001msn/direct/01/?href=http://get.live.com/messenger/overview


Back to: Top of message | Previous page | Main SAS-L page