Finding Equivalent Standards in Small Samples

Abstract

The process of resetting performance standards, with small samples, in different forms of an exam is statistically challenging as the estimates are often biased. Empirical information is therefore, often neglected and content experts reset the standard. In the current article, performance standards are set to a new form in by different methods: circle-arc equating and concurrent calibration with OPLM as an IRT model. The responses on the exam forms that are to be equated are simulated in different situations varying in sample size, test length, test difficulty and respondent’s abilities. The results demonstrate that even in small samples (50 subjects taking both tests), the IRT-method with OPLM as a model outperforms circle-arc equating when test difficulty and population ability interact.