Chronology Current Month Current Thread Current Date
[Year List] [Month List (current year)] [Date Index] [Thread Index] [Thread Prev] [Thread Next] [Date Prev] [Date Next]

Re: PDT Gain Factor



Gene:

You may want to ask this question on PhysLrnR-L; the original paper presenter
is probably there.

At the Anaheim AAPT meeting someone gave a talk on the g factor for the
PTD (FCI) exam. The traditional g factor is

What do you mean by PTD? I don't know this acronym. Please explain.


g = (posttest - pretest) / (30 - prestest)

where 30 is a perfect score. The problem will this g factor is that one
set of scores can provide a disproportional impact on the classes g-factor
average. This only happens when the pretest score is high and the postest
score is less high. For example, if the pretest score is 29 and the
posttest score is 26 the traditional g factor is -4. Since the maximum
possible g factor is +1, a g factor of -4 has a disproportionate impact on
the class average. The situation is even worse if the pretest score is a
perfect 30. Then, a less then perfect posttest score give a traditional g
factor of NAN (Not A Number) since the denominator is zero.

I get (26-29)/(30-29) = -3/1 = -3; but your point is taken. The real problem
is a ceiling effect here, the FCI is not well-designed to discriminate at
its uppermost bounds. Scores at 25+ are going to be less reliable; the test
is not designed to work well here. I would justify not using these data
on grounds that these scores are not instrument reliable. If this is one
student, they may know the FCI and not been focussed on the second testing.
Their scores do not accurately represent their mechanics ability at these high
numbers but rather their ability to stay focussed on the test taking.
I have had similar students (high-end scholarship types) in my data and have
excluded them for similar reasons.


Anyway, the speaker proposed an improved g factor that avoided such
disproportionate impacts. Unfortunately, I forgot it. Can
someone out with a better memory than mine tell me what it is?

I recall the presenter as being a male grad student, either from College
Park or U Wash (sorry I can do no better). His problems with the Hake gain
<g> all lay with cases that showed decreases on scores or extremely
unusual cases with floor and ceiling effects. The FCI just doesn't work
reliably at extrema like these, no standardized instrument does.


Gene

----------------------------------------------------------------
Eugene (Gene) P. Mosca Phone 410-293-6659 Office
Physics Department 410-267-0144 Home
572 Holloway Road 610-683-3597 Summer
U.S.Naval Academy Mailstop 9C, Michelson Rm 339
Annapolis MD 21402 email: mosca@nadn.navy.mil
----------------------------------------------------------------

Dan M

Dan MacIsaac, Assistant Professor of Physics and Astronomy, Northern AZ Univ
danmac@nau.edu http://purcell.phy.nau.edu PHYS-L list owner