Forget questions.  That's so western.  Publish the tasks they will have to
perform, of which they'll have to really know how to do stuff, and pepper them
with questions as a guide to what matters most.  The weight should be on
successfully accomplishing the dynamic implementation of system features in a
live environment.

David Weeks


On Mon, 12 Feb 2001, you wrote:
> From: Michael Dowling <[EMAIL PROTECTED]>
> 
> 
> : Please do elaborate.  It seems to me that making the items public allows
> for
> : greater numbers of editors and testers, since there is no longer any need
> to
> : keep them secret while doing so.
> 
> I don't have time to elaborate fully (nor spell-check) and this is a
> hypothetical discussion because of the other difficulties I outlined in my
> long previous email.  But, hypothetically, let's take one example.  What if
> I only memorized a few questions and they appear on my exam.  for those
> questions, my exam score reflects my Linux ability and some error
> (regardless of my ability, I got those items right).  And notice that the
> error is always "positive".  Under this model, there is no "downward" force
> so the net effect is always to get a little higher score.  So this would
> introduce a bias into the test scores. Even if it were small it would be
> bad, and I see no reason to assume it would always be small.  In fact, I
> think it would vary, some people would be much advantaged and others not at
> all.  This is a psychometrician's nightmare... tests should measure what you
> test as much as is possible excluding other factors.  A test publisher's job
> first and foremost, is to reduce bias and error in the scores of it's exams.
> 
> Furthernore, one cannot randomly select items unless they are equivalent.
> If there is variation in difficulty, for example, then random selection
> would give some people harder tests and other people easier tests.  But you
> would hold them to the same cut-score.  Does that seem fair?
> 
> No, it seems ametuerish.  What you have to do is to pre-calibrate the items
> so you know how easy or hard they are.  And that means intensive study of
> each item.. so an enormous pool would be enormously costly (in resources,
> not necssarily money).  I described this in my last post.
> 
> Besides, even if training companies were not involved, don't you think that
> publishing our items would turn our test into a test of one's ability to
> memorize (instead of on'e Linux skills)?  Isn't that exactly what we DO NOT
> want certification tests to do?  I certainly hear people complain about this
> influence all the time about all certification exams.  I think, again, it's
> our job to remove this component as much as possible.
> 
> Finally, for what purpose?  I don't think we are constraioned about who can
> provide items.  Didn't we decide that there was no prohibition?  So this is
> a solution seeking a problem.  I think it flows more from a sense that
> things should be open (like Linux).  I agree whole-heartedly with that
> attitude and I love open-source; that's why I volunteer for LPI.  But that
> analogy simply doesn't make for exams like the one we're discussing.
> 
> So those are a few quick thoughts.  I invite couter-arguments if I've missed
> something.
> 
> -Alan Mead
> 
> 
> --
> This message was sent from the lpi-examdev mailing list.
> Send `unsubscribe lpi-examdev' in the subject to [EMAIL PROTECTED] 
> to leave the list.

--
This message was sent from the lpi-examdev mailing list.
Send `unsubscribe lpi-examdev' in the subject to [EMAIL PROTECTED] 
to leave the list.

Reply via email to