Abstract:
This study examined the substantive and predictive validity facets of the University Entrance Examination for English Major (UEEEM) students. To that aim, 111 English major students were recruited to report their scores on each of the subtests of the test as well as their grade point average. Sequential multiple regressions and factor analysis were used in the analysis of the data. Results acknowledged a 2-factor structure underlying the test. Moreover, multiple regression analyses indicated that the presence of a large number of seemingly construct irrelevant items, Arabic, and theology, coupled with items with no unique contribution to variation to the response variable, the General English subtest, has compromised the predictive validity of the test. The paper concludes with the implications it carries for the test’s stakeholders, particularly its developers and score users.
Machine summary:
Moreover, multiple regression analyses indicated that the presence of a large number of seemingly construct irrelevant items, Arabic, and theology, coupled with items with no unique contribution to variation to the response variable, the General English subtest, has compromised the predictive validity of the test.
The reason for this high frequency is clear: The entire business of testing and assessment in both general and language education comes down to designing and administering valid instruments for the measurement of intended abilities or knowledge areas.
Consistent with this unified conceptualization of validity, this paper examines two validity facets, the predictive and substantive, of the University Entrance Examination for English Major (UEEEM), the national language test that is administered annually to screen candidates for undergraduate language programs at state universities in Iran.
Using a differential item functioning (DIF) approach, Barati, Ketabi, and Ahmadi (2006) investigated the general English module of the University Entrance Examination, administered to high school graduates of math, sciences, and humanities, who sit the test for admission to tertiary institutions.
The next research question is about the test modules that explain the observed variance in the participants’ GPAs as the response variable to be predicted by the specified set of explanatory variable: Special English, General English, Persian, Arabic, and Theology.
The three multiple regression analyses reported above evidently demonstrate that the UEEEM fails to explain a large portion of variance in the participants’ GPAs. It was also revealed that the two subtests of Special English and Persian contribute the largest amount of explanation to variations in the response variable.