From: Daniel Ginsburg (dginsburg@gmail.com)
Date: Sun May 15 2005 - 15:28:32 GMT-3
On 5/15/05, Scott Morris <swm@emanon.com> wrote:
> Whenever you take great math which involves a theorhetical number pulled
> from someplace where the sun doesn't shine, we end up with "statistics"!
> While yes, the truth is in the numbers, they can also be manipulated to tell
> any story you'd like.
>
> With the 'no partial credit' rule, we very often find people who "came
> really close" But assuming you have 20 point sections at 5 points each, and
> each point section has 10 items. If I successfully complete 9 of the 10 in
> EVERY section, technically I completed 90% of the exam correctly. However
> my score would be 0. So thinking you came close and really coming close are
> two differerent ideas!
>
I won't really argue that the average is higher or lower or equal to
50. I was only pointig out deficiency in Dillon's calculation when he
arbitrary assumes that average _is_ 50. Without such assumption his
calclulations are not very meaningful.
> Personally, I think time would be better spent figuring out how traffic
> shaping works, or tracking down whoever came up with the theory about token
> buckets so we can burn them in effigy. But if we want to analyze the crap
> out of a score report and postulate on the global woes about who does or
> doesn't pass, that's an individual's decision. Just not one (IMHO) that is
> well-placed.
>
> In the end, 80 points matter. And when people pass you never see your
> score. So you can believe you got 97 points all you want, but you REALLY
> may only have 80. Concentrate on passing. Score reports are meant to give
> you an indicator of what areas you need more work in...
>
I still wish they give score report when you pass. :)
> Scott
>
> -----Original Message-----
> From: nobody@groupstudy.com [mailto:nobody@groupstudy.com] On Behalf Of
> Daniel Ginsburg
> Sent: Sunday, May 15, 2005 1:46 PM
> To: ccielab@groupstudy.com
> Subject: Re: Dillon's open critique about CCIE lab exam
>
> On 5/14/05, Dillon Yang <dillony@gmail.com> wrote:
>
> [snip]
>
> > 1.1 Theoretic analysis
> > Regarding the scores of any exam, it is subjected to the normal
> > distribution. For normal distribution, the F(x) indicate the
> > probability from 0.0000 to 1.0000, the x is from 3.09 to 3.09, for
> > example, when x is 0, the F(x) that means all probability below 0 is
> > 0.5, and when x is 1.854, the F(x) will be about 0.9682. You can find
> > the value by a common statistical table.
> > Dillon can standardization change the scores that is from 0 to 100 to
> > the x in F(x) by the formula x=(S-u)/b. The u means the average of all
> > scores, and the b is the something of tolerance. Because the average
> > of a score must be 50, so Dillon get u equal 50, and Dillon have
> > 3.09=(100-50)/b, so the b equal 16.1813. Now Dillon can get that x
> > equals 1.854 when the S equals 80, 1.854=(80-50)/16.1813. That means
> > the 0.9682 of the total candidate will be rejected if Cisco stick to
> > its rule of 80. Dillon n other words, only 0.0318 of the total
> > candidates can pass the lab exam if the 80-pass rule is the truth.
>
> There's insufficiently grounded assumption here: you postulate that the
> average score is 50. I believe that it is more than that. Remember how often
> you hear "I almost nailed it, but missed only few points". I don't think
> these people are liars. If average is more than 50 then assuming normal
> distribution much more than 3.18% of attempts are successful.
>
> [snip]
>
> > 2 Quirky Wording
> > Maybe Cisco noticed that something abnormity, and adopted the
> > nonsensical wording. Cisco maybe believe that the changing wording can
> > hold out the cheating without essential modification. Yes, essential
> > modifications will be more expensive than only changing wording, but
> > we are not all the master degree of literature, even if we are not all
> > that english is his mother tongue. Remember? Dillon n the written
> > exam, the candidate will have more 30 minutes if his mother tongue is
> > not english while the american can get only 2 hours. Why not in lab
> > exam?
> > The wording is really efficient for those cheating, and for those that
> > not cheating, too. Did Cisco ever think about that if an engineer
> > designs or implements a network for his clients, his client maybe ask
> > unintelligible questions or requirements but he will explain it
> > throughout with common wording to help the engineer to finish the job.
> > Now, CCIE lab exam gradually becomes a english exam, not technique
> > exam, Dillon MHO.
>
> While I'm not a native speaker and my English is way too far from perfect I
> found wording of the exam clear enough to understand almost every task. When
> I wasn't sure I asked proctor who was very nice and answered most of my
> questions.
>
> --
> dg
> #14229
>
> _______________________________________________________________________
> Subscription information may be found at:
> http://www.groupstudy.com/list/CCIELab.html
>
>
-- dg #14229
This archive was generated by hypermail 2.1.4 : Fri Jun 03 2005 - 10:11:58 GMT-3