Comparison of Test Administration Procedures for Placement Decisions in a Mathematics Course

All tests were prepared from an item response theory calibrated item bank. The subjects were 90 volunteer students from three adult education schools. They were randomly assigned to one of six experimental groups to take two tests which differed in mode of administration. The results indicate that test performance was not differentially affected by the mode of administration and that the CAT always yielded more accurate ability estimates than the two other test administration procedures. The CAT was also found to be capable of making placement decisions with a test that was on average 24 % shorter.