| Summary: | The issue which has always been discussed by scholars in the area of language teaching and testing is whether the test is valid i.e. whether it tests what it is supposed to test and whether the test is reliable i.e. consistent in assessing the candidates. This research attempts to construct and to validate a placement test for new students at the Academy of Islamic Studies at the University of Malaya in Malaysia. The design of the text based on the syllabus at the Academy as well as the test specification, prepared during this research. Four sub-tests are constructed: Reading, Grammar, Writing and Dictation. To ensure the validity of the test, two analyses, internal and external, are conducted. The internal validity analysis is concerned with face and content validity. Three group of students from different levels of academic background and countries participated in the pilot study for the purpose of internal validity analysis. Modifications are made to some items of the sub-tests at the end of the pilot study. The external validity analysis is concerned with concurrent and predictive validity. The correlation coefficients (<I>r</I>) between the total mark of the sub-tests and one of the two measures for concurrent validity indicate that the relationships are moderate: between .40 and .60. As for predictive validity, the <I>r</I> between the sub-tests and the total mark of the final examination are between .60 and .64: a moderate relationship too. Another aspect of validity, construct validity analysis, is the topic of concern for this study too. The construct validity analysis, using Factor analysis method, groups the total score of the sub-tests into three factors namely Reading, Grammar, and Writing. The Dictation test is loaded in the Writing test. In the analysis of the reliability of the tests using the internal consistency method, the reliability coefficients (<I>r<SUB>xx </SUB></I>) for the sub-tests are very high: ranging between .87 and .90. The correlation analysis between the total score of the sub-tests also indicates a very high relationship: five correlation coefficients (<I>r</I>) are between .70 and .75 and only one correlation has the <I>r</I> of .69.
|
|---|
