Hi

On 5 Feb 2003, dennis roberts wrote:

> At 12:57 PM 2/5/2003, jim clark wrote:
> >I routinely raise grades just below boundaries to the next higher
> >level, not for institutional reasons, but rather for what I think
> >are valid statistical/measurement reasons (biased somewhat
> >perhaps by a desire to err on the side of the students).
> >
> >Or if you thought about a confidence interval around obtained
> >grades, the likelihood that the interval around 79 overlaps with
> >80 is much higher than that the interval around 78 overlaps with
> >80.
> >
> >So I see the practice as simply some acknowledgement that our
> >measures involve error (perhaps unlike jumping between roof-tops,
> >as mentioned in a previous post).
> 
> of course ... the problem with this is CI error band notion
> is ..... that, we could use this same argument to RAISE the
> cut scores too ... or round the students' scores DOWN
> 
> measurement error works BOTH ways and ... giving the benefit
> of the doubt to the student is not the same thing as
> accounting for measurement error
> 
> it is not really measurement error that drives us to DOWNWARD
> move the cut scores ... or round student scores UPward ...
> it's that we know that WE are not perfect and ... would
> rather add some tolerance to OUR decision making to save our
> hides such that ... IF a student were to come and challenge
> us ... we can point to the fact that we have moved either
> their score up or ... the cut score down ... hence trying to
> FEND off arguments and grade hassles

I'm not sure I fully appreciate the distinction between
"measurement error" and "not perfect" or "tolerance."  In any
case, for me, the measurement error view provides a response to
people who would argue that adjusting grades leads to an infinite
regress ... that is, 79 becomes 80, but 78 becomes 79 which could
then become 80, and so on.  The idea of a sampling distribution
around some underlying value provides a rationale for moving 79
and not 80, or 78 and 79 but not 77, or whatever set point,
depending on how much risk instructors are willing to take of
failing to raise a student whose grades come from the underlying
A distribution.

Best wishes
Jim

============================================================================
James M. Clark                          (204) 786-9757
Department of Psychology                (204) 774-4134 Fax
University of Winnipeg                  4L05D
Winnipeg, Manitoba  R3B 2E9             [EMAIL PROTECTED]
CANADA                                  http://www.uwinnipeg.ca/~clark
============================================================================

.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to