A Modified Regression Approach to the Problem of Comparing Two or More Groups with only One Paper in Common.

Many English Examination Boards will make some use of the differentiated paper method of assessment in the new scheme of 16+ examining, but it is not clear whether problems of comparability have been adequately solved. This article presents an analysis based on well-defined assumptions incorporated into a measurement model and utilising regression. Marks on the paper not taken by a group are estimated from their joint marks on the common paper and the paper specific to their group. The marks are then weighted using the measurement model and combined to produce a rank order over both the groups. In an appendix, the weighting method used is related to the more usual method of weighting adopted when combining scores. [ABSTRACT FROM AUTHOR]