Statistics in linguistic measurement

Spearman’s rank correlation coefficient ρ and Kendall’s τ have been applied in psychological measurement widely. In this thesis, we develop an alternative approach to calculate the correlation coefficient by considering the nature of the test items since the accuracy of the results of the multivariate analysis depends on the accuracy of the correlation coefficient provided. Therefore, there is a need to determine a correlation coefficient which is accurate and accommodate the properties of the data. Item response theory (IRT) is not only the measurement theory underlying many major tests such as GRE today, but also it has many important research applications. However, we are aware of no IRT programs available as part of commonly used statistical packages such as SAS and SPSS. We thus develop a SAS program to estimate the parameters of the unidimensional IRT model