Universitas Terbuka uses a multiple-choice format
test as the only means for assessing students'
achievement. Universitas Terbuka does not
statistically analyze its item collections although
item analysis is a crucial process in any objective
test development. In this study, I propose the
establishment of an item bank model at Universitas
Terbuka. The main task in developing the item bank
system is to carry out test item analyses. In this
study I analyzed test items using two approaches:
classical item analysis, and the Rasch Model.
The main features of classical item analysis
include p-value, discriminating power, and distractor
analysis. The purpose of doing this analysis was to
delete poor items and items were retained if r.pbis >
0.2 and all options were chosen by some students.
Furthermore, only one dLstractor was allowed to have a
positive r.pbis up to 0.05. Classical item analysis
showed only 38% of items on the December 1986 and 54%
of items on the May 1987 English examination were
satisfactory. Furthermore, 33% of items in December
1986 and 70% of items in the May 1987 Mathematics examination were satisfactory. The rest of the items were not suitable for assessing student achievement because they had poor characteristics.
The Rasch Model provides a different approach to
item analysis. This model is very useful for adjusting
all items in the collection onto a common difficulty
scale, especially when teachers want to create
different tests to measure the same objectives. In
this study a number of items common to the December
1986 and May 1987 examination were used to adjust all
the remaining items onto a common difficulty scale.
To benefit from the calibrated items, all items
need to be entered into an item bank in order that item
collections are easy to be retrieved. In the final
chapter of this study, I propose a systematic procedure
for creating, field-testing, analyzing, calibrating
items, and constructing an item bank to manage test
production.