絶對基準檢査 製作에 관한 問題点

Issues selected to the construction of Criterion-Referenced Test

본 논문은 절대기준검사가 그 검사 특성에 알맞게 특별히 제작된 제작 방법과 일치한다면 학습목표의 성취여부를 기본으로 하는 교육 프로그램에서는 그 효율성이 높을 것이라는 전제하에 영역기준검사를 제작하는데 있어서 문헌에 나타난 경험적 연구방법을 소개하고 그 문제점을 밝히고 있다.

To instruct a learner effectively, we should know what kinds of skills and how much of them a student has. The test measuring the skills should be consistent with instructional objectives. Criterion-referenced test is constructed for this purpose. (To say more correctly, we call such test a "domain-referenced test.") For the construction and evaluation of criterion-referenced test as well as interpretation methods related to the test results, various researchers have suggested many methods. However, the theories are not established yet. This article included introduction of empirical research methods for the construction of domain referenced test, and tried to present some problems. The problems were examined in four aspects such as : primarily, the writer of this article showed the methods defining the content domain. Os burn(1968) Bormuth(1970) and Hiverly(1968) constructed item generation rules form which item forms are derived. The contents included in the forms are the random samples which reflect the content domains. Secondly, construction methods for the items of domain-referenced test were introduced in two ways, that is, item construction by empirical data and item generation rules. The process of item construction by empirical materials are similar to that if item construction for norm-referenced test. However, item analysis should be suitable for criterion-referenced items. The method by Bormuth(1970) is time-consuming and his method is hardly applicable to many content areas. The method by Hively(1968) has similar limitations. It seems that Osburn's method(1969) may be easily applicable to more diverse areas. Thirdly, the concepts of measurement errors of criterion-referenced test can be examined in two ways : the estimation of domain score and instructional decisions for learners. Lord and Novick(1968) explained the concept of generic measurement error by defining it interms of linear equation. In the instructional decisions, accuracy and consistency are required. It is not necessarily true that being consistent means accuracy. It can be consistently inaccurate. Finally, the problems relevant to the concept of reliability and validity of criterion-referenced test are suggested. Classical meanings of reliability can not be applied to criterion-referenced measurement. Though many methods by different researchers are already introduced to measure that reliability of criterion-referenced test, however, enough empirial studies have not been published yet to show which method is better than the others. When item generation rules are used, content validity is guaranteed in some degrees. But it is desirable to examine the validity once more since it is not possible to define content universe completely.