To a or not to a: On the Use of the Total Score

For the sake of transparency, the use of the unweighted total score is demanded by society in many cases, especially in high-stakes situations such as exams. In the Rasch model, the total score is the sufficient statistic: all relevant information of the measurement is captured by the unweighted sum of the item scores. For this reason, many practitioners want to use the Rasch model. However, in many practical applications, the Rasch model does not fit, and the data is better described by a model that also uses a slope parameter. Although in these types of models, the total score is not the sufficient statistic; the unweighted item sum score can be used to compare candidates’ results on different equated tests. In a revaluation of the true-score equating procedure, we show how the benefits of using the better fitting model can be combined with the application of the total score in the context of equating cut-off scores. The advantages of the total scores are presented, and how the total score can be used also in case the Rasch model does not hold. An example is given to describe how the procedure works in practice. Finally, some reflections are given on the practical implications, meaning, and usefulness of the slope parameter, also known as the a-parameter.

In: Essays on Contemporary Psychometrics (Eds. Van der Ark, Emons & Meijer). Cham: Springer Verlag.