Abstract:

The purpose of this study is to equate Trends in International Mathematics and Science Study (TIMSS) mathematics subtest scores obtained from TIMSS 2011 to scores obtained from TIMSS 2007 form with different nonlinear observed score equating methods under Non-Equivalent Anchor Test (NEAT) design where common items are used to link two or more test forms. The ultimate goal is to determine whether different forms of mathematics tests that administered in different years with anchor (common) items caused any inequalities with respect to students. In addition, results obtained from chained and frequency estimation based on equipercentile equating methods were compared to four different methods (Tucker, Levine, Braun-Holland and chained) based on a new nonlinear equating approach called circle-arc equating in order to see which method is the most appropriate for equating these forms. The results of different nonlinear equating methods were compared with respect to Root Mean Squared Error (RMSE) index, mean of bootstrap standard errors (MBSE) and mean of bootstrap bias. Results indicates that TIMSS 2007 mathematics tests were easier than TIMSS 2011 mathematics across the score scale which indicates that results were biased against to students participated to TIMSS 2007. Moreover, equating methods based on nonlinear circle-arc equating outperformed the equipercentile equating methods and presmoothing decreased both standard error and bias associated with each method.

Keywords: