Improvement of Difficulty Parameter Equating base on Ability Parameter Distribution Adjustment

In automated test construction, one of the remaining issues is an inaccurate estimation of exam item parameters from different examinee ability distribution in linear equating. This work proposes a method to adjust the examinee's ability distribution to prevent carrying on error rates from the previous calibration to affect later procedures. By making the examinee's ability distribution in the minor' group to be similar to the examinee's ability distribution in the major distribution, the few errors are generated. From experiments, the proposed methods yield better results than those of baseline in terms of Root Mean Square Error in all test combinations.