Linking and anchoring techniques in test equating using the Rasch model

In analyzing computer assisted tests, the comparison between batches of examinees across different time frames is very important. In real testing situations, though great effort has been paid to construct parallel tests with the aim to ensure the fairness of the comparison between different cohorts using different tests, and overlap of test items in these different tests is often seen, an overall comparison is still not operational due to the existence of missing blocks in the data. In the traditional test theory (TTT), the standardization criterion is commonly used for the purpose of comparing the examinees’ achievement. However, the comparison is unarguable only when the examinees are considered from the same population. Using the Rasch model in the Item Response Theory (IRT), this paper presents the Linking and Anchoring techniques for comparing or equating tests. The advantages of these techniques are (1) the requirement that the examinees must be from same population is lifted and thus the comparison of examinee performance can be conducted across different academic levels; (2) the techniques are applicable even when the historic raw data are not available; and (3) the calibration of items using these techniques provides essential and comparable indexing reference for item banking.