From: FolkertsT at BARTONCCC.EDU (Folkerts, Timothy J)
Date: Wed Dec 15 19:17:23 2004
Brian Walcott had posted an interesting site with data mining and sta=
tistis of the US Senate at <http://www.ailab.si/aleks/Politics/> .=
=20
I was wondering if anyone knew of a similar analysis of FCI data. It=
seems only natural that numbers-oriented physicists would pursue suc=
h information, but searches of the archives and the web don't turn up=
much beyond calculations of gain at the class level. There are lost=
s of things one might want to know:
* which questions are missed most often?
* which questions do students improve most on?
* are there pairs of questions that students tend to perfom similarly=
on (e.g. do students who miss #4 also tend to miss #12)?
* do different techniques (lecture, peer instruction...) work best fo=
r certain questions?=20
I have a copy of Minitab which will do lots of interesting statisics,=
but I currently teach fairly small classes (<20 per year) so I don't=
have enough data to analyze.=20
So... If anyone is interested and has data, I am interested in crunch=
ing some numbers. Ideally you would have raw pre/post test data for =
a large number of students. Even more ideally, it would be possible =
to associate the pre/post data for each student (i.e. some sort of st=
udent ID # with each row of data). =20
I would post a summary of my findings here at Phys-L (the specific so=
urce of the data would be kept anonymous). I could also forward resu=
lts back to the sources of data. Ultimately, I can see turning this =
into a publication if there is something interesting that emergesd fr=
om the data, but I would get permission from any contributors before =
including their data in this sort of formal report.
Feel free to discuss the general idea on Phys-L or email data directl=
y to me at folkertst@bartonccc.edu . Any format that I can import i=
nto Excel would work. THANKS!