초록 열기/닫기 버튼

본 연구는 Rasch 모형을 활용하여 기술신용보증기금의 기술평가 시 평가자의 일관성 및 성향을 분석하였다. 이를 위해, 2005년 7월부터 2007년 9월까지 기술신용보증기금에서 기술평가가 수행된 13,343건을 대상으로 하였으며, 평가에 참여한 평가자는 1,411명이다. 기술평가는 총 45개 평가 문항으로 구성되어 있으며 이 중 33개 주관평가 문항을 대상으로 Rasch 모형에 적용하였다. 그 결과, 전체적인 평가의 적합도 검정결과 일관성을 유지하고 있으며 문항에서도 적정히 유지되고 있다. 하지만 평가자에 대해서는 평가자의 약 30% 이상이 관대하게 평가하고 있어 문제가 되고 있으며 전체 1411명 중 11명 0.8% 정도도 평가 일관성에 문제가 있어 이에 대한 개선이 필요하다.


This paper examines the level of the consistency and tendency for evaluators in using the technology rating data of KIBO(Korea Technology Credit Guarantee Fund). The KTRS (Kibo Technology rating system) rating data of KIBO have the 13,343 rating from July 2005 to August 2007 and evaluators conducted are 1141 analyst persons. Technology rating is consisted of the 45 rating items, projective 33 rating items of this items are applied to the Rasch model. The results of this paper are as follows. First, the examination result of fitness maintains the consistency of total model. Second, the consistency and tendency of items getting out of a criterion are one item each. Finally, as to the consistency and tendency of the evaluators, approximately 30% of all evaluators have been evaluated tolerantly, 11 evaluators of all evaluators have a problem with the consistency.


This paper examines the level of the consistency and tendency for evaluators in using the technology rating data of KIBO(Korea Technology Credit Guarantee Fund). The KTRS (Kibo Technology rating system) rating data of KIBO have the 13,343 rating from July 2005 to August 2007 and evaluators conducted are 1141 analyst persons. Technology rating is consisted of the 45 rating items, projective 33 rating items of this items are applied to the Rasch model. The results of this paper are as follows. First, the examination result of fitness maintains the consistency of total model. Second, the consistency and tendency of items getting out of a criterion are one item each. Finally, as to the consistency and tendency of the evaluators, approximately 30% of all evaluators have been evaluated tolerantly, 11 evaluators of all evaluators have a problem with the consistency.