Extending a Scale of Language Proficiency Using Concurrent Calibration and the Rasch Model

The ability of the Rasch model computer program BIGSTEPS to perform concurrent calibration and differences in the outcomes between vertical equating using a common item anchoring method and a common item concurrent calibration method was examined. The data for the investigation came from two Chinese language tests that researchers wished to equate: the extant Chinese Proficiency Test developed in 1984, and a new, lower-level version called the Preliminary Chinese Proficiency Test developed in 1991. Both tests are designed to evaluate the level of general proficiency in Chinese listening and reading comprehension attained by Americans and other English-speaking learners of Chinese. This paper demonstrates the ability of BIGSTEPS to vertically equate a scale of Chinese language proficiency using a concurrent calibration method. In studies on all the items based on the test data, only very minor differences were found between the item difficulty logit values produced by the two methods and those produced when no equating was involved. However, in studies involving only the common items, the concurrent calibration method was found to have a beneficial effect on the calibration of the common items. Users of BIGSTEPS are encouraged to employ the concurrent calibration method in test equating. Contains 4 references. (LB)