--- Forwarded Message from "Fuqiang Zhuo" <[log in to unmask]> ---
>Date: Tue, 21 Dec 2004 14:41:00 -0800 (PST)
>To: [log in to unmask]
>Subject: Question: Inter-Item Reliability Test
>From: "Fuqiang Zhuo" <[log in to unmask]>
Hi,
The Japanese placement exam computerized by me in Macromedia Director MX
2004 during the summer contains 250 multiple-choice items (A, B, C, D) in
five sections (Listening, Kanji Characters, Grammar, Vocabulary, and
Reading). Each section has different numbers of items which have been
unevenly divided into 3 levels: 1-2, 3-4, and 5-6. The program will
automatically jump to next section if a student does not meet the pre-
determined correct percentage at any level in any section. The speaking
and writing sections are not my concern here.
There is no reliability test analysis conducted previously by the
instructors, and we will definitely modify the computer program again to
improve the placement exam in the coming summer. Some day we may change
to a web-based placement exam. Therefore, we will need to do a
reliability test and other related analyses such as item difficulty and
discrimination. It is very easy for me to use SPSS or other statistics
software to perform statistical analyses. What I am concerned about is
that since many students did not have the chance to go through all the
levels and all the sections, i.e., all the 250 items, whether using all
the examinees and throwing the possible 250 items into analyses is
reasonable. Or should I analyze the collected data based on each level
and each section? Simply put, what is the best approach to do such an
inter-item reliability test based on the above situations?
Thank you very much for your suggestions.
Fuqiang Zhuo
Language Learning Center
UC Davis
***********************************************
LLTI is a service of IALLT, the International Association for
Language Learning, and The Consortium for Language Teaching and
Learning (http://consortium.dartmouth.edu).
Join IALLT at http://iallt.org.
Otmar Foelsche, LLTI-Editor ([log in to unmask])
***********************************************
|
