Since 1977 the SweSAT has been used as an instrument for selection in admissions to higher education. Parallel to the development of new versions of tests, extensive research and development work is undertaken to assess and assure the quality of the SweSAT. One aspect of the work with quality assurance is to examine to what extent there are sequential effects in the subtest Data Sufficiency DS.
All newly constructed items are tried out in connection with the regular tests. The day of the test consists of five test periods of 50 minutes each, one of which is the try out items. This means that all test-takers are doing two versions of one of the tests.
Consequently some test-takers will do two subtests of the same kind one after the other – where one is an ordinary test and one is a try-out test. The test-takers do not know which of these will contribute to their official test score and which one consists of new items that are being assessed and that will be used in future SweSATs. The try-out tests are randomly distributed over the 24 universities and colleges.
The aim of this study has been to examine to what extent there are sequential effects in the subtest Data Sufficiency, i.e. if the first DS subtest affects achievement on the second DS subtest when the two tests are distributed with only a short break in between.
The sequential effect has been analysed with the use of a structural equation model and an index that relates the results for different groups of test-takers to the result of all test-takers.
The results show a clear tendency – if two DS subtests are given in succession, with a short break of around 30-minutes in between tests, the average result of the second test will be higher than when the tests are not administered in succession. On average there is a difference of approximately one point out of a total of twenty-two. The index method shows that the test-takers at all the ten examined universities and colleges performed better on the second DS-subtest when the tests were administered in succession.
Umeå: Institutionen för beteendevetenskapliga mätningar, Umeå , 2007. , 21 p.