一般化線形モデル(GLM)によるG-TELPスコアからTOEICスコアの推定モデルの構築 : 長崎大学学生の2011年から2016年のデータから [in Japanese] Estimating the TOEIC Scores from the G-TELP Scores by the Generalized Linear Model : From the Data Obtained from Nagasaki University Students from 2011 to 2016 [in Japanese]
Access this Article
Search this Article
This article aims to estimate the TOEIC scores from the G-TELP (Level 3) scores with the data obtained from Nagasaki University students from 2011 to 2016. The problems with the previous estimation research lie in the inadequate fit and the use of the linear regression model, which assumes residuals being normally distributed. This study uses both the linear regression model and the Generalized Linear Model (GLM), which can handle categorical variables and is more flexible in the assumption on error structure. Both departments andentrance years are included as factors in the GLM to predict the TOEIC scores from the G-TELP scores. The results indicate that the estimation by the GLM is better overall to predict the TOEIC scores than the linear regression model, suggesting (a) departments and entrance years should be included in the model in estimating the TOEIC scores, (b) asis the case with the previous research, the estimated scores in the lowest or the highest score ranges are not so precise, and (c) the GLM is more appropriate in estimating the scores than the traditional linear regression model. Further research should be necessary that will take into account individual differences and/or time lag between the two tests.
- 長崎大学言語教育研究センター論集 = Journal of Center for Language Studies, Nagasaki University
長崎大学言語教育研究センター論集 = Journal of Center for Language Studies, Nagasaki University (6), 33-51, 2018-03