A Change Point Analysis Method Combining Response And Response Time To Detect Test Speededness

Test speededness is frequently observed in psychological and educational tests,which occurs when an examinee is aware there is not enough time for her/him to finish the test,then s/he speeds up responding to the items and inevitably leading to more mistakes.Since speededness undermines test validity and fairness,a suitable approach to handle it is urgently needed.Change point analysis(CPA),which is a flexible tool in statistical process control,can be used to detect speededness.However,the existing CPA methods only focus on response data,which usually have relatively low detection power and do not work satisfactorily in operational testing.We believe one promising way to improve these methods is to combine response data with response time data that can provides effective clues for speededness.Under this idea,the current study proposed a CPA method that combined the two data types to detect speededness.The performance of the proposed method was evaluated in the simulation study in comparison with the existing CPA methods.Results indicated that the proposed method outperformed the existing methods in power with type-I error rate well controlled,and could locate speeding point more accurately,which leaded to an effective cleansing of data and an improvement of parameter calibration.A real data of the Raven Advanced Progressive Matrices was analyzed to evaluate the empirical validity of the proposed method.Finally,we discussed the implications and limitations of the current study,as well as the future directions.