I had written, in part: > >... Even given take-home exams and a matter of weeks to do them, > >most papers are in fact turned in (in my experience) practically AT > >the deadline.
to which Dennis Roberts replied: > > well, there can be lots of reasons for this, not the least might be > the perception on the part of the students that if you say it is due > friday in class .... then turn it in friday IN class ... NOT before > (nor after) Sounds to me like a conjecture. I do not know whether my students ever had any such perception or not; I'm inclined to suspect not, but have no relevant data. > ... surely, you can't distinguish in this case between the student > who needed exactly that amount of time ... from one who had it done > 3 days earlier but, was not by your office until the deadline ... > from a student who really needed MORE time (yes ... > procrastinated!) but, turns it in since if he/she doesn't ... he/she > gets a zero Nope. Wasn't trying to. Why would one want to? In any case, you can't either, I think. > my point is: DATA AT THIS BUNCH UP POINT IS MURKY DATA A point on which we all may agree, supposing we had agreed on what "data" we had in mind. As far as it represents the performances exhibited by the students, it is not very murky at all. Of course, if one wants to infer what *abilities* the students have, that's a horse of another colour -- and degree of murkiness! -- entirely. > > Hard to see support for the idea that the distribution of > >time taken to complete an assignment (whether an exam or otherwise) is > >"artificially produced". Seems to me to be the nature of the beast: > >quite naturally produced. > > no ... i don't agree ... WHY would you think that most tend to bunch > up at that point? Same reason(s) you do, I should imagine. Most of 'em were implied in the earlier post. > there are at least 2 baselines we can visualize ... > > baseline 1 > > how long before you turn the test in given the 1 hour time limit > > baseline 2 > > how long before you turn the test in given that you have as much > time as you need Umm... When is that ever given? (And according to whose perception of "as long as you need"?) > i am not suggesting that the observable data does not form a > radically - skewed distribution ... i am just suggesting that the > "amount" of it's - skew is partly a function of the fact that you > insist that they turn in the test in no more than 60 minutes ... While acknowledging that such an insistence is usually in place in fact, I deny responsibility for it. That, in my experience, is a restriction imposed by the college or university, with which I am compelled to work. (As is the even sillier insistence that all papers be graded within 48 hours of the final examination. An effective recipe for lousy data.) > even GOOD students sometimes hold on to their tests as long as they > can ... I see your students are not unlike mine. Surprising? > what if the test above were 60 items ... which produces the > radically - skew ... given 1 hour > now, you cut the test to 30 items ... but, give the same 1 hour ... > ? you are not going to see the same degree of - skew ... that is for > sure ?Is it? Again, sounds like conjecture to me. Do you have supporting evidence for this proposition? (And where do you find 60 minutes for a test? Every college I've taught at has had either 50-minute classes (three times a week) or 75-minute classes (twice a week), and class tests are expected to be completed within an ordinary class period. For end-of-term tests an examination schedule is commonly published, but the exam times tend to be 2 hours, or 2.5, or 3; not 1.) > as long as we don't interpret this data as meaning ... "the amount > of time NEEDED to complete the test ... " ... i am ok with it ... A straw man. Has anyone attempted such an interpretation? > on a related matter ... what if one wanted to examine the > relationship between the amount of time needed to complete a test > AND the score ON the test? with a fixed time limit ... say like the > 1 hour above ... you do get this bunching effect near 60 minutes and > the distinctions between the times more or less vanishes .... > whereas we can be pretty sure that someone who turns it in in 20 > minutes (which someone will do) has spent less time than someone who > turned it in in 40 .... but, when you have 9 people who give it to > you at minute 60 ... one can't make the assumption that they equally > needed 60 minutes ... if the time limit were not a limit ... then, > one could examine the relationship between time needed and score ... > in a much cleaner way If you were to write "between time taken and score" I might agree, just for conversation. But "time needed"? I have no idea how to observe, let alone measure, such a thing; and if you claim you can, I'll be VERY skeptical of the claim. ----------------------------------------------------------------------- Donald F. Burrill [EMAIL PROTECTED] 56 Sebbins Pond Drive, Bedford, NH 03110 (603) 626-0816 . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
