awej 5 no.4 full issue 2014

AWEJ Volume.5 Number.4, 2014Fair, Reliable, Valid: Developing a Grammar Test UtilizingEfeotorTheMiketableOrrshows that the Cronbach Alpha person reliability value is 0.84, and the modelreliability is 0.83. Table 2 shows the item reliability value of 0.89. These results indicate that theestimated measures are reliable. This suggests that the test allows us to discriminate betweenexaminees based on their ability, and also discriminate between items based on their difficulty(Bond & Fox, 2001).Table 2.The Item Reliability ValueValidityThe second step in the process is to check for test validity. The meaning of validity in testing hasevolved over time. In 1954, the American Psychological Association purported four types ofvalidity: concurrent, construct, content and predictive. Predictive and concurrent validity werelater combined to form criterion-related validity (Smith, 2001). For Messick (1989, 1995),―validity is a unitary concept realized in construct validity and has six facets of content,substantive, structural, generalizability, external, and consequential‖ (as cited in Baghaei &Amrahi, 201, p.1052). It is this definition of validity that will form the basis of the analysis, withthe most important aspects discussed. The data was analysed using WINSTEPS Rasch Software.A glance at the item person matrix yields some preliminary observations possibly affectingvalidity, such as overfit.Arab World English JournalISSN: 2229-9327www.awej.org213