The Impact of Population Shift on Equating: An Empirical Study

• In Event: Studies Investigating Demographic and Other Factors Impacting Assessment Development, Decision Making, Policy, and Practice

Abstract

Recently, population shift occurred in our testing program. As a result, the proportion of examinees from each subgroup is different between the new and the reference forms. It is necessary for us to examine the impact of population shift on equating. Two non-linear equating methods under the classical equating framework were conducted on sampled data created from real data by simple random sampling with replacement. These sampled data were manipulated to have different levels of population shifts between the new and the reference forms. The results showed that equating bias became unacceptably large when the proportion difference for a certain subgroup between the new and reference forms was larger than or equal to 30%.

Authors

• Yanxuan Qu, ETS

• Youhua Wei, ETS

• Rick Morgan, ETS