An Empirical Study on Assessment of Item-Person Statistics and Reliability Using Classical Test Theory Measurement Methods

Abstract

Recent developments in statistics and psychology supported an increase in quality of measurement in the educational area.The present study empirically examined the characteristics of item and person statistics of mathematics category of SchoolOlympiad Examination (SOE) within the Classical Test Theory (CTT) measurement frameworks. When constructing tests, especiallyin examining reliability of a test, CTT approach gives opportunity to obtain required targets easily. In this study, variousstatistics are used to judge the quality of the items. Mainly, item discrimination, item difficulty, item-total inter correlations arefound to clarify real aspect of data. Before determining internal consistency, unidimensionality of mathematics examination isanalyzed. In this process, Principal Component Analysis is implemented and using item component correlation, behaviors ofitems are detected from different point of view. Internal consistency is examined by Cronbach’s Alpha. Standard error of measurementis calculated to identify confidence interval around an observed score. As a conclusion, reliability of the mathematicspart of the SOE is proved by CTT assumptions.