The impact of re-sitting on the fairness of grading
One consideration in allowing re-sitting is that examinations are not perfectly reliable instruments. On any occasion a candidate may perform slightly better or worse depending on a number of chance factors related to their preparation and the test content. From this perspective it may seem fair to allow candidates to re-sit to compensate for the unreliability inherent in measuring ability through examinations.
The extent to which candidates’ performance is likely to differ on notional parallel tests and the impact this will have on their grades can be measured through measures of classification accuracy. If a test were perfectly reliable then all candidates would be classified into their ‘true’ grade. As the test becomes less reliable the grades they receive are likely to vary. They may perform worse than their true grade (a false negative) or better (a false positive). It is possible to simulate the performance of candidates on parallel tests using measures of the ability of the candidates and the difficulties of the items