A Calibrated Item Bank for Computerized Adaptive Testing in Measuring Science TIMSS Performance

The current assessment is demanding for a more personalised and less-time consuming testing environment. Computer Adaptive Testing (CAT) is seemed as a more effective alternative testing method in comparison to conventional test in meeting the current standard of assessment. This research reports on the calibration of the released Grade 8 Science objective items in Trends in International Mathematics and Science Study (TIMSS) 2003-2015 based on Rasch model framework to be used in CAT as an alternative testing tool in a low stakes test. Concurrent common item equating method was used in linking and equating sets of items. Five test sets were produced consisting of 20 unique items and 10 common items in each set. The unique items were available only in a single set while common items were available in the two-consecutive set of items. The sets were administered through Paper and Pencil test to Form 2 (Grade 8) students who had been selected through a purposive sampling method from secondary schools in the northern part of Malaysia. The fit analysis, polarity analysis, unidimensionality analysis, item measure and Person-Map-Item were conducted. The analysis produced 122 calibrated items which meet the Rasch’s requirements and were suitable to be used in CAT.